BRD Pathogens - BNA Manuscript
Brittany L Morgan Bustamante
12/09/2022
1. P. multocida
Farm level factors
First analysis is an assessment of farm-level factors on P. multocida resistance/susceptibile MIC interpretations. Variables at this level under consideration include type of calf rearing facility (single source, multisource), whether the farm practices onsite milking, vaccines or biologicals administered to calves on site, colostrum administration at arrival to multisource facility, colostrum pasteurization, hospital milk pasteurization, bull calves comingled with heifers, milk mass medication, grain mass medication, water mass medication, feed lane cleaning method in weaned pens (lagoon flush, scrape, clean water flush).
Any variable with a cell size less than 10 will be recategorized, if possible. If it isn’t possible to recategorize to meet a minimum count of 10, the variable will be removed from analysis.
1.1 Data Wrangling
pmult.df <- read_csv("pmult.csv", show_col_types = FALSE)
# connect calf treatment variables to pmult df
pmult.raw <- left_join(pmult.df, calf.trx, by = "calf_ID")
# make variables factors, add labels
pmult.factor = pmult.raw
pmult.factor$season <- factor(pmult.raw$season, order = FALSE,
levels = c(1, 2),
labels = c("summer", "winter"))
pmult.factor$brd <- factor(pmult.raw$brd, order = FALSE,
levels = c(0, 1),
labels = c("neg", "pos"))
pmult.factor$cef <- factor(pmult.raw$cef, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$dan <- factor(pmult.raw$dan, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$enr <- factor(pmult.raw$enr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$flr <- factor(pmult.raw$flr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$gam <- factor(pmult.raw$gam, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$gen <- factor(pmult.raw$gen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$pen <- factor(pmult.raw$pen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$spc <- factor(pmult.raw$spc, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$tet <- factor(pmult.raw$tet, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$tild <- factor(pmult.raw$tild, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$tilm <- factor(pmult.raw$tilm, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$tul <- factor(pmult.raw$tul, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
pmult.factor$source_no <- factor(pmult.raw$source_no, order = FALSE,
levels = c(1, 2, 3),
labels = c("single", "low multi", "high multi"))
pmult.factor$onsite_milking <- factor(pmult.raw$onsite_milking, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$resp_vacc <- factor(pmult.raw$resp_vacc, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$salmonella <- factor(pmult.raw$salmonella, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$first_def <- factor(pmult.raw$first_def, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$bovishield <- factor(pmult.raw$bovishield, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$pyramid <- factor(pmult.raw$pyramid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$titanium <- factor(pmult.raw$titanium, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$pinkeye <- factor(pmult.raw$pinkeye, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$clostrid <- factor(pmult.raw$clostrid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$colost_source <- factor(pmult.raw$colost_source, order = FALSE,
levels = c(1, 2),
labels = c("p", "dp"))
pmult.factor$colost_past <- factor(pmult.raw$colost_past, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$colost_again <- factor(pmult.raw$colost_again, order = FALSE,
levels = c(0, 1, 99),
labels = c("no", "yes", "-99"))
pmult.factor$hosp_milk_pastuerized <- factor(pmult.raw$hosp_milk_pastuerized, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$action <- factor(pmult.raw$action, order = FALSE,
levels = c(1, 2),
labels = c("screen", "premium"))
pmult.factor$bull_calves <- factor(pmult.raw$bull_calves, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$milk_mass_med <- factor(pmult.raw$milk_mass_med, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mass_pen_exp <- factor(pmult.raw$mass_pen_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mass_tet_exp <- factor(pmult.raw$mass_tet_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mass_corr_exp <- factor(pmult.raw$mass_corr_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mass_rum_exp <- factor(pmult.raw$mass_rum_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mass_sul_exp <- factor(pmult.raw$mass_sul_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$feed_lane_cleaning <- factor(pmult.raw$feed_lane_cleaning, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("scrape", "lagoon", "clean", "sl"))
pmult.factor$dust_mgmt <- factor(pmult.raw$dust_mgmt, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$mort_rate <- factor(pmult.raw$mort_rate, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$pen_trx_rate <- factor(pmult.raw$pen_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$cef_trx_rate <- factor(pmult.raw$cef_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$mac_trx_rate <- factor(pmult.raw$mac_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$res_nuf_trx_rate <- factor(pmult.raw$res_nuf_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$tet_inj_use <- factor(pmult.raw$tet_inj_use, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
pmult.factor$sul_trx_rate <- factor(pmult.raw$sul_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$bay_trx_rate <- factor(pmult.raw$bay_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
pmult.factor$Farm <- factor(pmult.raw$Farm, order = FALSE,
levels = c(1, 2, 3, 4, 5, 6),
labels = c("A", "B", "C", "D", "E", "F"))
pmult.factor$last_treat <- factor(pmult.raw$last_treat, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days", "no trx"))
pmult.factor$no_classes <- factor(pmult.raw$no_classes, order = FALSE,
levels = c(0, 1, 2, 3),
labels = c("0", "1", "2 or more", "2 or more"))
pmult.factor$total_abx <- factor(pmult.raw$total_abx, order = FALSE,
levels = c(0, 1, 2),
labels = c("0", "1 to 2", "3 or more"))
pmult.factor$days_since_pen <- factor(pmult.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
pmult.factor$days_since_ceph <- factor(pmult.raw$days_since_ceph, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
pmult.factor$days_since_mac <- factor(pmult.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
pmult.factor$days_since_phen <- factor(pmult.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
pmult.factor$days_since_tet <- factor(pmult.raw$days_since_tet, order = FALSE,
levels = c(2, 3, 4),
labels = c("16-60 days", ">=60 days", "no trx"))
pmult.factor$days_since_sul <- factor(pmult.raw$days_since_sul, order = FALSE,
levels = c(3, 4),
labels = c(">=60 days", "no trx"))
pmult.factor$days_since_flr <- factor(pmult.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
# glimpse(pmult.factor)# analytic data frame for all farms
pmult.all <- pmult.factor %>%
select(-"srp", -"entervene", -"pen_trx_rate", -"cef_trx_rate",
-"mac_trx_rate", -"res_nuf_trx_rate", -"brd",
-"tet_inj_use", -"sul_trx_rate", -"bay_trx_rate", -"mort_rate",
- "calf_ID", -"Farm_ID", -"cef", -"tet", -"total_abx", -"last_treat",
-"days_since_pen", -"days_since_ceph", -"days_since_mac",
-"days_since_phen", -"days_since_tet", -"days_since_sul",
-"days_since_flr")
# check cell counts - All Farms
summary(pmult.all) # gen, pen, below 10## Farm season dan enr flr gam gen pen spc
## A:13 summer:71 S:55 S:55 S:59 S: 41 S:139 S:142 S:110
## B:24 winter:74 R:90 R:90 R:86 R:104 R: 6 R: 3 R: 35
## C:21
## D:34
## E:24
## F:29
## tild tilm tul source_no onsite_milking resp_vacc first_def
## S: 36 S: 33 S:120 single :63 no : 34 no : 24 no :116
## R:109 R:112 R: 25 low multi :37 yes:111 yes:121 yes: 29
## high multi:45
##
##
##
## salmonella bovishield pyramid titanium pinkeye clostrid colost_source
## no :87 no :84 no :95 no :111 no : 37 no :111 p :121
## yes:58 yes:61 yes:50 yes: 34 yes:108 yes: 34 dp: 24
##
##
##
##
## colost_past colost_again hosp_milk_pastuerized action bull_calves
## no : 24 no :37 no : 24 screen :124 no :69
## yes:121 yes:45 yes:121 premium: 21 yes:76
## -99:63
##
##
##
## milk_mass_med mass_pen_exp mass_tet_exp mass_rum_exp mass_corr_exp
## no :79 no :132 no : 29 no : 13 no :76
## yes:66 yes: 13 yes:116 yes:132 yes:69
##
##
##
##
## mass_sul_exp feed_lane_cleaning dust_mgmt no_classes
## no :53 scrape:34 no : 24 0 :45
## yes:92 lagoon:53 yes:121 1 :56
## clean :34 2 or more:44
## sl :24
##
## # remove variables
pmultall.analyze <- as.data.frame(pmult.all) %>%
select(-"gen", -"pen", -"colost_again")
# check
summary(pmultall.analyze)## Farm season dan enr flr gam spc tild tilm
## A:13 summer:71 S:55 S:55 S:59 S: 41 S:110 S: 36 S: 33
## B:24 winter:74 R:90 R:90 R:86 R:104 R: 35 R:109 R:112
## C:21
## D:34
## E:24
## F:29
## tul source_no onsite_milking resp_vacc first_def salmonella
## S:120 single :63 no : 34 no : 24 no :116 no :87
## R: 25 low multi :37 yes:111 yes:121 yes: 29 yes:58
## high multi:45
##
##
##
## bovishield pyramid titanium pinkeye clostrid colost_source colost_past
## no :84 no :95 no :111 no : 37 no :111 p :121 no : 24
## yes:61 yes:50 yes: 34 yes:108 yes: 34 dp: 24 yes:121
##
##
##
##
## hosp_milk_pastuerized action bull_calves milk_mass_med mass_pen_exp
## no : 24 screen :124 no :69 no :79 no :132
## yes:121 premium: 21 yes:76 yes:66 yes: 13
##
##
##
##
## mass_tet_exp mass_rum_exp mass_corr_exp mass_sul_exp feed_lane_cleaning
## no : 29 no : 13 no :76 no :53 scrape:34
## yes:116 yes:132 yes:69 yes:92 lagoon:53
## clean :34
## sl :24
##
##
## dust_mgmt no_classes
## no : 24 0 :45
## yes:121 1 :56
## 2 or more:44
##
##
## Create an allowlist from season to farm to account for clustering of farm samples in each season. This will be the same allowlist used for all analyses
wl = data.frame(from = c("season"), to = c("Farm"))Create a blocklist of risk factor variables to season or farm, this accounts for multilevel structure. The blocklist will change for each analysis depending on which variables are included
# create levels
pmult.bl.all = tiers2blacklist(list("season", "Farm",
c("dan", "enr", "flr", "gam", "spc", "tild", "tilm", "tul",
"source_no", "onsite_milking", "resp_vacc",
"first_def", "salmonella", "bovishield", "pyramid",
"titanium", "pinkeye", "clostrid",
"colost_source", "colost_past", "action",
"hosp_milk_pastuerized", "bull_calves", "milk_mass_med",
"mass_pen_exp", "mass_tet_exp", "mass_rum_exp", "mass_corr_exp",
"mass_sul_exp", "feed_lane_cleaning", "dust_mgmt", "no_classes")))
pmult.bl.all = rbind(pmult.bl.all, c("season"))
pmult.bl.all## from to
## [1,] "Farm" "season"
## [2,] "dan" "season"
## [3,] "enr" "season"
## [4,] "flr" "season"
## [5,] "gam" "season"
## [6,] "spc" "season"
## [7,] "tild" "season"
## [8,] "tilm" "season"
## [9,] "tul" "season"
## [10,] "source_no" "season"
## [11,] "onsite_milking" "season"
## [12,] "resp_vacc" "season"
## [13,] "first_def" "season"
## [14,] "salmonella" "season"
## [15,] "bovishield" "season"
## [16,] "pyramid" "season"
## [17,] "titanium" "season"
## [18,] "pinkeye" "season"
## [19,] "clostrid" "season"
## [20,] "colost_source" "season"
## [21,] "colost_past" "season"
## [22,] "action" "season"
## [23,] "hosp_milk_pastuerized" "season"
## [24,] "bull_calves" "season"
## [25,] "milk_mass_med" "season"
## [26,] "mass_pen_exp" "season"
## [27,] "mass_tet_exp" "season"
## [28,] "mass_rum_exp" "season"
## [29,] "mass_corr_exp" "season"
## [30,] "mass_sul_exp" "season"
## [31,] "feed_lane_cleaning" "season"
## [32,] "dust_mgmt" "season"
## [33,] "no_classes" "season"
## [34,] "dan" "Farm"
## [35,] "enr" "Farm"
## [36,] "flr" "Farm"
## [37,] "gam" "Farm"
## [38,] "spc" "Farm"
## [39,] "tild" "Farm"
## [40,] "tilm" "Farm"
## [41,] "tul" "Farm"
## [42,] "source_no" "Farm"
## [43,] "onsite_milking" "Farm"
## [44,] "resp_vacc" "Farm"
## [45,] "first_def" "Farm"
## [46,] "salmonella" "Farm"
## [47,] "bovishield" "Farm"
## [48,] "pyramid" "Farm"
## [49,] "titanium" "Farm"
## [50,] "pinkeye" "Farm"
## [51,] "clostrid" "Farm"
## [52,] "colost_source" "Farm"
## [53,] "colost_past" "Farm"
## [54,] "action" "Farm"
## [55,] "hosp_milk_pastuerized" "Farm"
## [56,] "bull_calves" "Farm"
## [57,] "milk_mass_med" "Farm"
## [58,] "mass_pen_exp" "Farm"
## [59,] "mass_tet_exp" "Farm"
## [60,] "mass_rum_exp" "Farm"
## [61,] "mass_corr_exp" "Farm"
## [62,] "mass_sul_exp" "Farm"
## [63,] "feed_lane_cleaning" "Farm"
## [64,] "dust_mgmt" "Farm"
## [65,] "no_classes" "Farm"
## [66,] "season" "season"1.2 BNA
Will learn network structure using Hill-Climbing (HC) learning algorithm and Akaike Information Criterion scoring procedure. The HC algorithm is a hueristic search that explores DAGs by single-arc addition and removal. Will test network structure sensitivity to learning algorithm chosen by also running TABU algorithm and comparing results.
# all farms hc structure learning
pmultfarms.hc <- hc(pmultall.analyze, blacklist = pmult.bl.all, whitelist = wl, score = "aic")
# check score
score(pmultfarms.hc, pmultall.analyze)## [1] -1132.98# plot customization
pmult.a <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultfarms.hc))
graph::nodeRenderInfo(pmult.a) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.a)
# all farms tabu structure learning
pmultfarms.tabu <- tabu(pmultall.analyze, blacklist = pmult.bl.all,
whitelist = wl, score = "aic")
# check score
score(pmultfarms.tabu, pmultall.analyze)## [1] -1132.98# plot customization
pmult.b <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultfarms.tabu))
graph::nodeRenderInfo(pmult.b) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.b)
graphviz.compare(pmultfarms.hc, pmultfarms.tabu, shape = "ellipse",
main= c("P.mult Farms HC DAG", "P.mult Farms Tabu DAG"))
The DAGs are the same! Indicating the structure is not sensitive to the learning algorithm chosen.
Now we’ll bootstrap the DAG to identify potentially spurious associations. This process runs 10,000 bootstrap samples and prunes edges that do not appear in at least 50% of the structures discovered in each sample. We’ll again use the HC algorithm and the AIC scoring procedure.
# bootstrap
pmultfarm.boot = boot.strength(pmultall.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = pmult.bl.all,
whitelist = wl, score = "aic"))
# take arcs with strength 50% and return average consensus network
pmultfarm.avg = averaged.network(pmultfarm.boot, threshold = 0.5)
# graph customization
pmult.c <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultfarm.avg))
graph::nodeRenderInfo(pmult.c) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.c)
Compare the averaged DAG with the single run DAG. The red lines are the edges that were pruned by bootstrapping.
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(pmultfarm.avg, pmultfarms.hc, shape = "ellipse",
main= c("Pmult Farm Level Risk Factors Averaged DAG",
"Pmult Farm Level Risk Factors Single DAG"))
Individual level factors
The same analysis above will be run for individual level factors. Factors under consideration for analysis are individual antibiotic treatments, what class, how long it has been sense the calf was treated, total number of antibiotic drugs the calf has received, and when the last treatment of any antibiotic was.
1.3 Data Wrangling
# analytic data frame for individual risk factors
pmult.ind <- pmult.factor %>%
select("Farm", "season", "brd", "dan", "enr", "flr", "gam", "pen", "spc",
"tet", "tild", "tilm", "tul", "total_abx", "last_treat",
"days_since_pen", "days_since_ceph", "days_since_mac", "days_since_phen",
"days_since_tet", "days_since_sul", "days_since_flr")
# check for missingness
summary(aggr(pmult.ind))
##
## Missings per variable:
## Variable Count
## Farm 0
## season 0
## brd 0
## dan 0
## enr 0
## flr 0
## gam 0
## pen 0
## spc 0
## tet 0
## tild 0
## tilm 0
## tul 0
## total_abx 24
## last_treat 24
## days_since_pen 24
## days_since_ceph 24
## days_since_mac 24
## days_since_phen 24
## days_since_tet 24
## days_since_sul 24
## days_since_flr 24
##
## Missings in combinations of variables:
## Combinations Count Percent
## 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0 121 83.44828
## 0:0:0:0:0:0:0:0:0:0:0:0:0:1:1:1:1:1:1:1:1:1 24 16.55172# check cell counts - All Farms
summary(pmult.ind) # -pen, - tet, days_since_pen(<60 days, >=60 days, no treatment), days_since_ceph(<60 days, >=60 days, no treatment), days_since_mac(<60 days, >=60 days, no treatment), days_since_phen(<60 days, >=60 days, no treatment), days_since_tet(<60 days, >=60 days, no treatment), days_since_flr(<60 days, >=60 days, no treatment)## Farm season brd dan enr flr gam pen spc
## A:13 summer:71 neg:76 S:55 S:55 S:59 S: 41 S:142 S:110
## B:24 winter:74 pos:69 R:90 R:90 R:86 R:104 R: 3 R: 35
## C:21
## D:34
## E:24
## F:29
## tet tild tilm tul total_abx last_treat
## S: 0 S: 36 S: 33 S:120 0 :21 <16 days :19
## R:145 R:109 R:112 R: 25 1 to 2 :34 16-60 days:33
## 3 or more:66 >=60 days :48
## NA's :24 no trx :21
## NA's :24
##
## days_since_pen days_since_ceph days_since_mac days_since_phen
## <16 days : 7 <16 days : 3 <16 days : 5 <16 days : 9
## 16-60 days:10 16-60 days: 7 16-60 days: 7 16-60 days:18
## >=60 days :10 >=60 days : 10 >=60 days :27 >=60 days :42
## no trx :94 no trx :101 no trx :82 no trx :52
## NA's :24 NA's : 24 NA's :24 NA's :24
##
## days_since_tet days_since_sul days_since_flr
## 16-60 days: 2 >=60 days: 20 <16 days : 2
## >=60 days :23 no trx :101 16-60 days:12
## no trx :96 NA's : 24 >=60 days :29
## NA's :24 no trx :78
## NA's :24
## # remove variables
pmultind2 <- as.data.frame(pmult.ind) %>%
select(-"pen", -"tet")
# check
summary(pmultind2)## Farm season brd dan enr flr gam spc tild
## A:13 summer:71 neg:76 S:55 S:55 S:59 S: 41 S:110 S: 36
## B:24 winter:74 pos:69 R:90 R:90 R:86 R:104 R: 35 R:109
## C:21
## D:34
## E:24
## F:29
## tilm tul total_abx last_treat days_since_pen
## S: 33 S:120 0 :21 <16 days :19 <16 days : 7
## R:112 R: 25 1 to 2 :34 16-60 days:33 16-60 days:10
## 3 or more:66 >=60 days :48 >=60 days :10
## NA's :24 no trx :21 no trx :94
## NA's :24 NA's :24
##
## days_since_ceph days_since_mac days_since_phen days_since_tet
## <16 days : 3 <16 days : 5 <16 days : 9 16-60 days: 2
## 16-60 days: 7 16-60 days: 7 16-60 days:18 >=60 days :23
## >=60 days : 10 >=60 days :27 >=60 days :42 no trx :96
## no trx :101 no trx :82 no trx :52 NA's :24
## NA's : 24 NA's :24 NA's :24
##
## days_since_sul days_since_flr
## >=60 days: 20 <16 days : 2
## no trx :101 16-60 days:12
## NA's : 24 >=60 days :29
## no trx :78
## NA's :24
## # recategorize
pmultind2$days_since_pen <- factor(pmult.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
pmultind2$days_since_ceph <- factor(pmult.raw$days_since_ceph, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
pmultind2$days_since_mac <- factor(pmult.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
pmultind2$days_since_phen <- factor(pmult.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
pmultind2$days_since_tet <- factor(pmult.raw$days_since_tet, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treatment", "treatment",
"treatment", "no treatment"))
pmultind2$days_since_flr <- factor(pmult.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
# remove Farm E
pmultind2$Farm <- factor(pmult.raw$Farm, order = FALSE,
levels = c(1, 2, 3, 4, 6),
labels = c("A", "B", "C", "D", "F"))
summary(pmultind2)## Farm season brd dan enr flr gam spc tild
## A :13 summer:71 neg:76 S:55 S:55 S:59 S: 41 S:110 S: 36
## B :24 winter:74 pos:69 R:90 R:90 R:86 R:104 R: 35 R:109
## C :21
## D :34
## F :29
## NA's:24
## tilm tul total_abx last_treat days_since_pen
## S: 33 S:120 0 :21 <16 days :19 <60 days :17
## R:112 R: 25 1 to 2 :34 16-60 days:33 >=60 days :10
## 3 or more:66 >=60 days :48 no treatment:94
## NA's :24 no trx :21 NA's :24
## NA's :24
##
## days_since_ceph days_since_mac days_since_phen
## <60 days : 10 <60 days :12 <60 days :27
## >=60 days : 10 >=60 days :27 >=60 days :42
## no treatment:101 no treatment:82 no treatment:52
## NA's : 24 NA's :24 NA's :24
##
##
## days_since_tet days_since_sul days_since_flr
## treatment :25 >=60 days: 20 <60 days :14
## no treatment:96 no trx :101 >=60 days :29
## NA's :24 NA's : 24 no treatment:78
## NA's :24
##
## pmultind.analyze <- na.omit(pmultind2)
# check
summary(pmultind.analyze)## Farm season brd dan enr flr gam spc tild tilm
## A:13 summer:61 neg:63 S:51 S:51 S:49 S:37 S:88 S:32 S:29
## B:24 winter:60 pos:58 R:70 R:70 R:72 R:84 R:33 R:89 R:92
## C:21
## D:34
## F:29
## tul total_abx last_treat days_since_pen days_since_ceph
## S:98 0 :21 <16 days :19 <60 days :17 <60 days : 10
## R:23 1 to 2 :34 16-60 days:33 >=60 days :10 >=60 days : 10
## 3 or more:66 >=60 days :48 no treatment:94 no treatment:101
## no trx :21
##
## days_since_mac days_since_phen days_since_tet days_since_sul
## <60 days :12 <60 days :27 treatment :25 >=60 days: 20
## >=60 days :27 >=60 days :42 no treatment:96 no trx :101
## no treatment:82 no treatment:52
##
##
## days_since_flr
## <60 days :14
## >=60 days :29
## no treatment:78
##
## # create levels
pmult.bl.ind = tiers2blacklist(list("season", "Farm",
c("brd", "dan", "enr", "flr", "gam", "spc", "tild", "tilm",
"tul", "total_abx", "last_treat", "days_since_pen",
"days_since_ceph", "days_since_mac", "days_since_phen",
"days_since_tet", "days_since_sul", "days_since_flr")))
pmult.bl.ind = rbind(pmult.bl.ind, c("season"))
pmult.bl.ind## from to
## [1,] "Farm" "season"
## [2,] "brd" "season"
## [3,] "dan" "season"
## [4,] "enr" "season"
## [5,] "flr" "season"
## [6,] "gam" "season"
## [7,] "spc" "season"
## [8,] "tild" "season"
## [9,] "tilm" "season"
## [10,] "tul" "season"
## [11,] "total_abx" "season"
## [12,] "last_treat" "season"
## [13,] "days_since_pen" "season"
## [14,] "days_since_ceph" "season"
## [15,] "days_since_mac" "season"
## [16,] "days_since_phen" "season"
## [17,] "days_since_tet" "season"
## [18,] "days_since_sul" "season"
## [19,] "days_since_flr" "season"
## [20,] "brd" "Farm"
## [21,] "dan" "Farm"
## [22,] "enr" "Farm"
## [23,] "flr" "Farm"
## [24,] "gam" "Farm"
## [25,] "spc" "Farm"
## [26,] "tild" "Farm"
## [27,] "tilm" "Farm"
## [28,] "tul" "Farm"
## [29,] "total_abx" "Farm"
## [30,] "last_treat" "Farm"
## [31,] "days_since_pen" "Farm"
## [32,] "days_since_ceph" "Farm"
## [33,] "days_since_mac" "Farm"
## [34,] "days_since_phen" "Farm"
## [35,] "days_since_tet" "Farm"
## [36,] "days_since_sul" "Farm"
## [37,] "days_since_flr" "Farm"
## [38,] "season" "season"1.4 BNA
# individual factors hc structure learning
pmultind.hc <- hc(pmultind.analyze, blacklist = pmult.bl.ind,
whitelist = wl, score = "aic")
# plot customization
pmult.c <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultind.hc))
graph::nodeRenderInfo(pmult.c) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.c)
# individual factors tabu structure learning
pmultind.tabu <- tabu(pmultind.analyze, blacklist = pmult.bl.ind,
whitelist = wl,
score = "aic")
# plot customization
pmult.d <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultind.tabu))
graph::nodeRenderInfo(pmult.d) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.d)
graphviz.compare(pmultind.hc, pmultind.tabu, shape = "ellipse",
main= c("Pmult Individual Risk Factors HC DAG",
"Pmult Individual Risk Factors Tabu DAG"))
# bootstrap
pmultind.boot = boot.strength(pmultind.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = pmult.bl.ind,
whitelist = wl, score = "aic"))
# take arcs with strength of 50% and return average consensus network
pmultind.avg = averaged.network(pmultind.boot, threshold = 0.5)
# graph customization
pmult.e <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(pmultind.avg))
graph::nodeRenderInfo(pmult.e) <- list(fontsize=60)
Rgraphviz::renderGraph(pmult.e)
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(pmultind.avg, pmultind.hc, shape = "ellipse",
main= c("Pmult Individual Level Risk Factors Averaged DAG",
"Pmult Individual Level Risk Factors Single DAG"))
1.5 Parameter Learning
Farm level query
# fit the parameters - bayesian method
pcmult.bayes = bn.fit(pmultfarm.avg, pmultall.analyze, method = "bayes")
# probability table
pcmult.bayes##
## Bayesian network parameters
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.08508159 0.09507830
## B 0.25291375 0.08165548
## C 0.15501166 0.13534676
## D 0.22494172 0.24272931
## E 0.14102564 0.18903803
## F 0.14102564 0.25615213
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.489726 0.510274
##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## tild
## dan S R
## S 0.93835616 0.19406393
## R 0.06164384 0.80593607
##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## dan
## enr S R
## S 0.995495495 0.002762431
## R 0.004504505 0.997237569
##
## Parameters of node flr (multinomial distribution)
##
## Conditional probability table:
## S R
## 0.4075342 0.5924658
##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D E F
## S 0.38607595 0.08620690 0.05118110 0.82195122 0.16896552 0.03714286
## R 0.61392405 0.91379310 0.94881890 0.17804878 0.83103448 0.96285714
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## dan
## spc S R
## S 0.4189189 0.9640884
## R 0.5810811 0.0359116
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , onsite_milking = no
##
## gam
## tild S R
## S 0.180000000 0.004424779
## R 0.820000000 0.995575221
##
## , , onsite_milking = yes
##
## gam
## tild S R
## S 0.996453901 0.001639344
## R 0.003546099 0.998360656
##
##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.910958904 0.002283105
## R 0.089041096 0.997716895
##
## Parameters of node tul (multinomial distribution)
##
## Conditional probability table:
##
## spc
## tul S R
## S 0.8891403 0.6267606
## R 0.1108597 0.3732394
##
## Parameters of node source_no (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## source_no A B C D E
## single 0.004219409 0.002298851 0.002624672 0.996747967 0.002298851
## low multi 0.991561181 0.995402299 0.002624672 0.001626016 0.002298851
## high multi 0.004219409 0.002298851 0.994750656 0.001626016 0.995402299
## Farm
## source_no F
## single 0.996190476
## low multi 0.001904762
## high multi 0.001904762
##
## Parameters of node onsite_milking (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## onsite_milking scrape lagoon clean sl
## no 0.996350365 0.002347418 0.003649635 0.005154639
## yes 0.003649635 0.997652582 0.996350365 0.994845361
##
## Parameters of node resp_vacc (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## resp_vacc scrape lagoon clean sl
## no 0.003649635 0.002347418 0.003649635 0.994845361
## yes 0.996350365 0.997652582 0.996350365 0.005154639
##
## Parameters of node first_def (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## first_def A B C D E
## no 0.993670886 0.996551724 0.996062992 0.997560976 0.996551724
## yes 0.006329114 0.003448276 0.003937008 0.002439024 0.003448276
## Farm
## first_def F
## no 0.002857143
## yes 0.997142857
##
## Parameters of node salmonella (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## salmonella A B C D E
## no 0.006329114 0.003448276 0.003937008 0.997560976 0.996551724
## yes 0.993670886 0.996551724 0.996062992 0.002439024 0.003448276
## Farm
## salmonella F
## no 0.997142857
## yes 0.002857143
##
## Parameters of node bovishield (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bovishield A B C D E
## no 0.006329114 0.003448276 0.996062992 0.997560976 0.003448276
## yes 0.993670886 0.996551724 0.003937008 0.002439024 0.996551724
## Farm
## bovishield F
## no 0.997142857
## yes 0.002857143
##
## Parameters of node pyramid (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pyramid A B C D E F
## no 0.993670886 0.996551724 0.003937008 0.997560976 0.996551724 0.002857143
## yes 0.006329114 0.003448276 0.996062992 0.002439024 0.003448276 0.997142857
##
## Parameters of node titanium (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## titanium scrape lagoon clean sl
## no 0.996350365 0.997652582 0.003649635 0.994845361
## yes 0.003649635 0.002347418 0.996350365 0.005154639
##
## Parameters of node pinkeye (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pinkeye A B C D E F
## no 0.993670886 0.003448276 0.003937008 0.002439024 0.996551724 0.002857143
## yes 0.006329114 0.996551724 0.996062992 0.997560976 0.003448276 0.997142857
##
## Parameters of node clostrid (multinomial distribution)
##
## Conditional probability table:
##
## titanium
## clostrid no yes
## no 0.997757848 0.007246377
## yes 0.002242152 0.992753623
##
## Parameters of node colost_source (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## colost_source no yes
## p 0.010204082 0.997942387
## dp 0.989795918 0.002057613
##
## Parameters of node colost_past (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## colost_past no yes
## no 0.989795918 0.002057613
## yes 0.010204082 0.997942387
##
## Parameters of node hosp_milk_pastuerized (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## hosp_milk_pastuerized no yes
## no 0.989795918 0.002057613
## yes 0.010204082 0.997942387
##
## Parameters of node action (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## action A B C D E
## screen 0.993670886 0.996551724 0.003937008 0.997560976 0.996551724
## premium 0.006329114 0.003448276 0.996062992 0.002439024 0.003448276
## Farm
## action F
## screen 0.997142857
## premium 0.002857143
##
## Parameters of node bull_calves (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bull_calves A B C D E
## no 0.006329114 0.996551724 0.996062992 0.002439024 0.996551724
## yes 0.993670886 0.003448276 0.003937008 0.997560976 0.003448276
## Farm
## bull_calves F
## no 0.002857143
## yes 0.997142857
##
## Parameters of node milk_mass_med (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## milk_mass_med A B C D E
## no 0.006329114 0.996551724 0.996062992 0.997560976 0.003448276
## yes 0.993670886 0.003448276 0.003937008 0.002439024 0.996551724
## Farm
## milk_mass_med F
## no 0.002857143
## yes 0.997142857
##
## Parameters of node mass_pen_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_pen_exp A B C D E
## no 0.006329114 0.996551724 0.996062992 0.997560976 0.996551724
## yes 0.993670886 0.003448276 0.003937008 0.002439024 0.003448276
## Farm
## mass_pen_exp F
## no 0.997142857
## yes 0.002857143
##
## Parameters of node mass_tet_exp (multinomial distribution)
##
## Conditional probability table:
##
## first_def
## mass_tet_exp no yes
## no 0.002145923 0.991525424
## yes 0.997854077 0.008474576
##
## Parameters of node mass_rum_exp (multinomial distribution)
##
## Conditional probability table:
##
## mass_pen_exp
## mass_rum_exp no yes
## no 0.001886792 0.981481481
## yes 0.998113208 0.018518519
##
## Parameters of node mass_corr_exp (multinomial distribution)
##
## Conditional probability table:
##
## bull_calves
## mass_corr_exp no yes
## no 0.003597122 0.996732026
## yes 0.996402878 0.003267974
##
## Parameters of node mass_sul_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_sul_exp A B C D E
## no 0.006329114 0.003448276 0.003937008 0.002439024 0.996551724
## yes 0.993670886 0.996551724 0.996062992 0.997560976 0.003448276
## Farm
## mass_sul_exp F
## no 0.997142857
## yes 0.002857143
##
## Parameters of node feed_lane_cleaning (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## feed_lane_cleaning A B C D E
## scrape 0.990506329 0.001724138 0.994094488 0.001219512 0.001724138
## lagoon 0.003164557 0.994827586 0.001968504 0.001219512 0.001724138
## clean 0.003164557 0.001724138 0.001968504 0.996341463 0.001724138
## sl 0.003164557 0.001724138 0.001968504 0.001219512 0.994827586
## Farm
## feed_lane_cleaning F
## scrape 0.001428571
## lagoon 0.995714286
## clean 0.001428571
## sl 0.001428571
##
## Parameters of node dust_mgmt (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## dust_mgmt no yes
## no 0.989795918 0.002057613
## yes 0.010204082 0.997942387
##
## Parameters of node no_classes (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## no_classes A B C D E
## 0 0.763713080 0.126436782 0.002624672 0.030894309 0.995402299
## 1 0.232067511 0.374712644 0.664041995 0.382113821 0.002298851
## 2 or more 0.004219409 0.498850575 0.333333333 0.586991870 0.002298851
## Farm
## no_classes F
## 0 0.241904762
## 1 0.584761905
## 2 or more 0.173333333# conditional probabilities of AMD
pcmult.bayes$gam##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D E F
## S 0.38607595 0.08620690 0.05118110 0.82195122 0.16896552 0.03714286
## R 0.61392405 0.91379310 0.94881890 0.17804878 0.83103448 0.96285714pcmult.bayes$tild##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , onsite_milking = no
##
## gam
## tild S R
## S 0.180000000 0.004424779
## R 0.820000000 0.995575221
##
## , , onsite_milking = yes
##
## gam
## tild S R
## S 0.996453901 0.001639344
## R 0.003546099 0.998360656pcmult.bayes$tilm##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.910958904 0.002283105
## R 0.089041096 0.997716895pcmult.bayes$dan##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## tild
## dan S R
## S 0.93835616 0.19406393
## R 0.06164384 0.80593607pcmult.bayes$enr##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## dan
## enr S R
## S 0.995495495 0.002762431
## R 0.004504505 0.997237569pcmult.bayes$spc##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## dan
## spc S R
## S 0.4189189 0.9640884
## R 0.5810811 0.0359116set.seed(111)
cpquery(pcmult.bayes, (gam =="R"), tild == "R")## [1] 0.9484033cpquery(pcmult.bayes, (gam =="R"), tild == "S")## [1] 0.008209539cpquery(pcmult.bayes, (tild =="R"), gam == "R")## [1] 0.9976157cpquery(pcmult.bayes, (tild =="R"), gam == "S")## [1] 0.1315699set.seed(112)
cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "yes")## [1] 0.6855354cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "no")## [1] 0.9600666cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "yes")## [1] 0.3161341cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "no")## [1] 0.03434951set.seed(113)
cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "yes" & feed_lane_cleaning == "scrape")## [1] 0.75cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "no" & feed_lane_cleaning == "scrape")## [1] 0.9666244cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "yes" & feed_lane_cleaning == "scrape")## [1] 0.3636364cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "no" & feed_lane_cleaning == "scrape")## [1] 0.0356832set.seed(114)
cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "yes" & feed_lane_cleaning == "lagoon")## [1] 0.9413061cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "no" & feed_lane_cleaning == "lagoon")## [1] 0.875cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "yes" & feed_lane_cleaning == "lagoon")## [1] 0.05970967cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "no" & feed_lane_cleaning == "lagoon")## [1] 0.125set.seed(115)
cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "yes" & feed_lane_cleaning == "clean")## [1] 0.1745962cpquery(pcmult.bayes, (tild == "R"), onsite_milking == "no" & feed_lane_cleaning == "clean")## [1] 0.7142857cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "yes" & feed_lane_cleaning == "clean")## [1] 0.8199748cpquery(pcmult.bayes, (tild == "S"), onsite_milking == "no" & feed_lane_cleaning == "clean")## [1] 0.1666667Individual level query
# fit the parameters - bayesian method
pc.bayes = bn.fit(pmultind.avg, pmultind.analyze, method = "bayes")
# probability table
pc.bayes##
## Bayesian network parameters
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.09918699 0.11735537
## B 0.29430894 0.10082645
## C 0.18048780 0.16694215
## D 0.26178862 0.29917355
## F 0.16422764 0.31570248
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.5040984 0.4959016
##
## Parameters of node brd (multinomial distribution)
##
## Conditional probability table:
## neg pos
## 0.5204918 0.4795082
##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## , , tild = S
##
## Farm
## dan A B C D F
## S 0.954545455 0.023809524 0.500000000 0.998220641 0.954545455
## R 0.045454545 0.976190476 0.500000000 0.001779359 0.045454545
##
## , , tild = R
##
## Farm
## dan A B C D F
## S 0.004132231 0.047511312 0.428909953 0.008196721 0.393238434
## R 0.995867769 0.952488688 0.571090047 0.991803279 0.606761566
##
##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## dan
## enr S R
## S 0.995145631 0.003546099
## R 0.004854369 0.996453901
##
## Parameters of node flr (multinomial distribution)
##
## Conditional probability table:
##
## , , season = summer
##
## Farm
## flr A B C D F
## S 0.008196721 0.058011050 0.274774775 0.934782609 0.103960396
## R 0.991803279 0.941988950 0.725225225 0.065217391 0.896039604
##
## , , season = winter
##
## Farm
## flr A B C D F
## S 0.147887324 0.500000000 0.103960396 0.997237569 0.316753927
## R 0.852112676 0.500000000 0.896039604 0.002762431 0.683246073
##
##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D F
## S 0.38636364 0.08677686 0.05188679 0.82163743 0.03767123
## R 0.61363636 0.91322314 0.94811321 0.17836257 0.96232877
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## , , dan = S
##
## Farm
## spc A B C D F
## S 0.954545455 0.045454545 0.005494505 0.286476868 0.995867769
## R 0.045454545 0.954545455 0.994505495 0.713523132 0.004132231
##
## , , dan = R
##
## Farm
## spc A B C D F
## S 0.995867769 0.954545455 0.830578512 0.991803279 0.997076023
## R 0.004132231 0.045454545 0.169421488 0.008196721 0.002923977
##
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## gam
## tild S R
## S 0.86000000 0.00295858
## R 0.14000000 0.99704142
##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.900000000 0.002793296
## R 0.100000000 0.997206704
##
## Parameters of node tul (multinomial distribution)
##
## Conditional probability table:
##
## , , spc = S
##
## Farm
## tul A B C D F
## S 0.690839695 0.771493213 0.995049505 0.996453901 0.963917526
## R 0.309160305 0.228506787 0.004950495 0.003546099 0.036082474
##
## , , spc = R
##
## Farm
## tul A B C D F
## S 0.500000000 0.023809524 0.004504505 0.997512438 0.500000000
## R 0.500000000 0.976190476 0.995495495 0.002487562 0.500000000
##
##
## Parameters of node total_abx (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## total_abx A B C D F
## 0 0.762626263 0.126721763 0.003144654 0.031189084 0.242009132
## 1 to 2 0.156565657 0.168044077 0.380503145 0.352826511 0.276255708
## 3 or more 0.080808081 0.705234160 0.616352201 0.615984405 0.481735160
##
## Parameters of node last_treat (multinomial distribution)
##
## Conditional probability table:
##
## total_abx
## last_treat 0 1 to 2 3 or more
## <16 days 0.003906250 0.206310680 0.182160804
## 16-60 days 0.003906250 0.264563107 0.363065327
## >=60 days 0.003906250 0.526699029 0.453517588
## no trx 0.988281250 0.002427184 0.001256281
##
## Parameters of node days_since_pen (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_pen A B C D F
## <60 days 0.005050505 0.126721763 0.663522013 0.001949318 0.002283105
## >=60 days 0.005050505 0.374655647 0.050314465 0.001949318 0.002283105
## no treatment 0.989898990 0.498622590 0.286163522 0.996101365 0.995433790
##
## Parameters of node days_since_ceph (multinomial distribution)
##
## Conditional probability table:
## <60 days >=60 days no treatment
## 0.08469945 0.08469945 0.83060109
##
## Parameters of node days_since_mac (multinomial distribution)
##
## Conditional probability table:
##
## last_treat
## days_since_mac <16 days 16-60 days >=60 days no trx
## <60 days 0.264069264 0.213032581 0.001727116 0.003921569
## >=60 days 0.160173160 0.243107769 0.333333333 0.003921569
## no treatment 0.575757576 0.543859649 0.664939551 0.992156863
##
## Parameters of node days_since_phen (multinomial distribution)
##
## Conditional probability table:
##
## , , last_treat = <16 days
##
## season
## days_since_phen summer winter
## <60 days 0.881278539 0.497942387
## >=60 days 0.004566210 0.300411523
## no treatment 0.114155251 0.201646091
##
## , , last_treat = 16-60 days
##
## season
## days_since_phen summer winter
## <60 days 0.644768856 0.188630491
## >=60 days 0.060827251 0.560723514
## no treatment 0.294403893 0.250645995
##
## , , last_treat = >=60 days
##
## season
## days_since_phen summer winter
## <60 days 0.001801802 0.001658375
## >=60 days 0.563963964 0.638474295
## no treatment 0.434234234 0.359867330
##
## , , last_treat = no trx
##
## season
## days_since_phen summer winter
## <60 days 0.003436426 0.004566210
## >=60 days 0.003436426 0.004566210
## no treatment 0.993127148 0.990867580
##
##
## Parameters of node days_since_tet (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_tet A B C D F
## treatment 0.007575758 0.004132231 0.004716981 0.587719298 0.174657534
## no treatment 0.992424242 0.995867769 0.995283019 0.412280702 0.825342466
##
## Parameters of node days_since_sul (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_sul A B C D F
## >=60 days 0.007575758 0.004132231 0.004716981 0.587719298 0.003424658
## no trx 0.992424242 0.995867769 0.995283019 0.412280702 0.996575342
##
## Parameters of node days_since_flr (multinomial distribution)
##
## Conditional probability table:
## <60 days >=60 days no treatment
## 0.1174863 0.2404372 0.6420765pc.bayes$flr##
## Parameters of node flr (multinomial distribution)
##
## Conditional probability table:
##
## , , season = summer
##
## Farm
## flr A B C D F
## S 0.008196721 0.058011050 0.274774775 0.934782609 0.103960396
## R 0.991803279 0.941988950 0.725225225 0.065217391 0.896039604
##
## , , season = winter
##
## Farm
## flr A B C D F
## S 0.147887324 0.500000000 0.103960396 0.997237569 0.316753927
## R 0.852112676 0.500000000 0.896039604 0.002762431 0.683246073pc.bayes$gam##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D F
## S 0.38636364 0.08677686 0.05188679 0.82163743 0.03767123
## R 0.61363636 0.91322314 0.94811321 0.17836257 0.96232877pc.bayes$tild##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## gam
## tild S R
## S 0.86000000 0.00295858
## R 0.14000000 0.99704142pc.bayes$tilm##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.900000000 0.002793296
## R 0.100000000 0.997206704pc.bayes$dan##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## , , tild = S
##
## Farm
## dan A B C D F
## S 0.954545455 0.023809524 0.500000000 0.998220641 0.954545455
## R 0.045454545 0.976190476 0.500000000 0.001779359 0.045454545
##
## , , tild = R
##
## Farm
## dan A B C D F
## S 0.004132231 0.047511312 0.428909953 0.008196721 0.393238434
## R 0.995867769 0.952488688 0.571090047 0.991803279 0.606761566pc.bayes$enr##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## dan
## enr S R
## S 0.995145631 0.003546099
## R 0.004854369 0.996453901pc.bayes$spc##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## , , dan = S
##
## Farm
## spc A B C D F
## S 0.954545455 0.045454545 0.005494505 0.286476868 0.995867769
## R 0.045454545 0.954545455 0.994505495 0.713523132 0.004132231
##
## , , dan = R
##
## Farm
## spc A B C D F
## S 0.995867769 0.954545455 0.830578512 0.991803279 0.997076023
## R 0.004132231 0.045454545 0.169421488 0.008196721 0.002923977pc.bayes$tul##
## Parameters of node tul (multinomial distribution)
##
## Conditional probability table:
##
## , , spc = S
##
## Farm
## tul A B C D F
## S 0.690839695 0.771493213 0.995049505 0.996453901 0.963917526
## R 0.309160305 0.228506787 0.004950495 0.003546099 0.036082474
##
## , , spc = R
##
## Farm
## tul A B C D F
## S 0.500000000 0.023809524 0.004504505 0.997512438 0.500000000
## R 0.500000000 0.976190476 0.995495495 0.002487562 0.500000000set.seed(115)
cpquery(pc.bayes, (tul =="R"), spc == "R")## [1] 0.4290794cpquery(pc.bayes, (tul =="R"), spc == "S")## [1] 0.1120944cpquery(pc.bayes, (tul =="S"), spc == "R")## [1] 0.5466461cpquery(pc.bayes, (tul =="S"), spc == "S")## [1] 0.8861281set.seed(116)
cpquery(pc.bayes, (flr =="R"), season == "summer")## [1] 0.6734533cpquery(pc.bayes, (flr =="R"), season == "winter")## [1] 0.5231815cpquery(pc.bayes, (flr =="S"), season == "summer")## [1] 0.3252704cpquery(pc.bayes, (flr =="S"), season == "winter")## [1] 0.47598082. M. haemolytica
The same analysis above will be run for the remaining two respiratory pathogens.
Farm level factors
2.1 Data Wrangling
# import data
mhaem.df <- read_csv("MHAEM.csv", show_col_types = FALSE)
# glimpse(mhaem.df)# connect calf treatment variables to mhaem df
mhaem.df <- mhaem.df %>%
mutate(calf_ID = ID) %>%
select(-"ID")
mhaem.raw <- left_join(mhaem.df, calf.trx, by = "calf_ID")
# make variables factors, add labels
mhaem.factor = mhaem.raw
mhaem.factor$season <- factor(mhaem.raw$season, order = FALSE,
levels = c(1, 2),
labels = c("summer", "winter"))
mhaem.factor$brd <- factor(mhaem.raw$brd, order = FALSE,
levels = c(0, 1),
labels = c("neg", "pos"))
mhaem.factor$cef <- factor(mhaem.raw$cef, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$dan <- factor(mhaem.raw$dan, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$enr <- factor(mhaem.raw$enr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$flr <- factor(mhaem.raw$flr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$gam <- factor(mhaem.raw$gam, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$gen <- factor(mhaem.raw$gen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$pen <- factor(mhaem.raw$pen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$spc <- factor(mhaem.raw$spc, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$tet <- factor(mhaem.raw$tet, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$tild <- factor(mhaem.raw$tild, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$tilm <- factor(mhaem.raw$tilm, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$tul <- factor(mhaem.raw$tul, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
mhaem.factor$source_no <- factor(mhaem.raw$source_no, order = FALSE,
levels = c(1, 2, 3),
labels = c("single", "low multi", "high multi"))
mhaem.factor$onsite_milking <- factor(mhaem.raw$onsite_milking, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$resp_vacc <- factor(mhaem.raw$resp_vacc, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$first_def <- factor(mhaem.raw$first_def, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$salmonella <- factor(mhaem.raw$salmonella, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$bovishield <- factor(mhaem.raw$bovishield, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$pyramid <- factor(mhaem.raw$pyramid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$titanium <- factor(mhaem.raw$titanium, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$pinkeye <- factor(mhaem.raw$pinkeye, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$clostrid <- factor(mhaem.raw$clostrid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$colost_source <- factor(mhaem.raw$colost_source, order = FALSE,
levels = c(1, 2),
labels = c("p", "dp"))
mhaem.factor$colost_past <- factor(mhaem.raw$colost_past, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$colost_again <- factor(mhaem.raw$colost_again, order = FALSE,
levels = c(0, 1, 99),
labels = c("no", "yes", "-99"))
mhaem.factor$hosp_milk_pastuerized <- factor(mhaem.raw$hosp_milk_pastuerized,
order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$action <- factor(mhaem.raw$action, order = FALSE,
levels = c(1, 2),
labels = c("screen", "premium"))
mhaem.factor$bull_calves <- factor(mhaem.raw$bull_calves, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$milk_mass_med <- factor(mhaem.raw$milk_mass_med, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mass_pen_exp <- factor(mhaem.raw$mass_pen_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mass_tet_exp <- factor(mhaem.raw$mass_tet_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mass_corr_exp <- factor(mhaem.raw$mass_corr_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mass_rum_exp <- factor(mhaem.raw$mass_rum_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mass_sul_exp <- factor(mhaem.raw$mass_sul_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$feed_lane_cleaning <- factor(mhaem.raw$feed_lane_cleaning, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("scrape", "lagoon", "clean", "sl"))
mhaem.factor$dust_mgmt <- factor(mhaem.raw$dust_mgmt, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$mort_rate <- factor(mhaem.raw$mort_rate, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$pen_trx_rate <- factor(mhaem.raw$pen_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$cef_trx_rate <- factor(mhaem.raw$cef_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$mac_trx_rate <- factor(mhaem.raw$mac_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$res_nuf_trx_rate <- factor(mhaem.raw$res_nuf_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$tet_inj_use <- factor(mhaem.raw$tet_inj_use, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
mhaem.factor$sul_trx_rate <- factor(mhaem.raw$sul_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$bay_trx_rate <- factor(mhaem.raw$bay_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
mhaem.factor$Farm <- factor(mhaem.raw$Farm, order = FALSE,
levels = c(1, 2, 3, 4, 5, 6),
labels = c("A", "B", "C", "D", "E", "F"))
mhaem.factor$last_treat <- factor(mhaem.raw$last_treat, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))
mhaem.factor$no_classes <- factor(mhaem.raw$no_classes, order = FALSE,
levels = c(0, 1, 2, 3),
labels = c("0", "1", "2", "3 or more"))
mhaem.factor$total_abx <- factor(mhaem.raw$total_abx, order = FALSE,
levels = c(0, 1, 2),
labels = c("0", "1 to 2", "3 or more"))
mhaem.factor$days_since_pen <- factor(mhaem.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))
mhaem.factor$days_since_ceph <- factor(mhaem.raw$days_since_ceph, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))
mhaem.factor$days_since_mac <- factor(mhaem.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))
mhaem.factor$days_since_phen <- factor(mhaem.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))
mhaem.factor$days_since_tet <- factor(mhaem.raw$days_since_tet, order = FALSE,
levels = c(2, 3, 4),
labels = c("16-60 days", ">=60 days", "no trx"))
mhaem.factor$days_since_sul <- factor(mhaem.raw$days_since_sul, order = FALSE,
levels = c(3, 4),
labels = c(">=60 days", "no trx"))
mhaem.factor$days_since_flr <- factor(mhaem.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days",
">=60 days", "no trx"))# analytic data frame for all farms
mhaem.all <- mhaem.factor %>%
select(-"srp", - "entervene", -"pen_trx_rate", -"brd",
-"cef_trx_rate", -"mac_trx_rate", -"res_nuf_trx_rate",
-"tet_inj_use", -"sul_trx_rate", -"bay_trx_rate", -"mort_rate",
- "calf_ID", -"Farm_ID", -"total_abx",
-"last_treat", -"days_since_pen", -"days_since_ceph",
-"days_since_mac", -"days_since_phen", -"days_since_tet",
-"days_since_sul", -"days_since_flr")
# check for missingness
summary(aggr(mhaem.all))
##
## Missings per variable:
## Variable Count
## Farm 0
## season 0
## cef 0
## dan 0
## enr 0
## flr 0
## gam 0
## gen 0
## pen 0
## spc 0
## tet 0
## tild 0
## tilm 0
## tul 0
## source_no 0
## onsite_milking 0
## resp_vacc 0
## salmonella 0
## first_def 0
## bovishield 0
## pyramid 0
## titanium 0
## pinkeye 0
## clostrid 0
## colost_source 0
## colost_past 0
## colost_again 0
## hosp_milk_pastuerized 0
## action 0
## bull_calves 0
## milk_mass_med 0
## mass_pen_exp 0
## mass_tet_exp 0
## mass_rum_exp 0
## mass_corr_exp 0
## mass_sul_exp 0
## feed_lane_cleaning 0
## dust_mgmt 0
## no_classes 0
##
## Missings in combinations of variables:
## Combinations
## 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0
## Count Percent
## 119 100# check cell counts - All Farms
summary(mhaem.all) # cef, gen, resp_vacc, colost_source, colost_past, hosp_milk_pastuerized, feed_lane_cleaning(sl), dust_mgmt, no_classes(3 or more) below 10## Farm season cef dan enr flr gam gen pen spc
## A:23 summer:60 S:119 S:36 S:34 S:79 S:75 S:112 S:74 S:100
## B:26 winter:59 R: 0 R:83 R:85 R:40 R:44 R: 7 R:45 R: 19
## C:17
## D:20
## E: 9
## F:24
## tet tild tilm tul source_no onsite_milking resp_vacc
## S: 16 S:55 S:43 S:84 single :44 no :40 no : 9
## R:103 R:64 R:76 R:35 low multi :49 yes:79 yes:110
## high multi:26
##
##
##
## salmonella first_def bovishield pyramid titanium pinkeye clostrid
## no :53 no :95 no :61 no :78 no :99 no :32 no :99
## yes:66 yes:24 yes:58 yes:41 yes:20 yes:87 yes:20
##
##
##
##
## colost_source colost_past colost_again hosp_milk_pastuerized action
## p :110 no : 9 no :32 no : 9 screen :102
## dp: 9 yes:110 yes:43 yes:110 premium: 17
## -99:44
##
##
##
## bull_calves milk_mass_med mass_pen_exp mass_tet_exp mass_rum_exp mass_corr_exp
## no :52 no :63 no :96 no :24 no :23 no :67
## yes:67 yes:56 yes:23 yes:95 yes:96 yes:52
##
##
##
##
## mass_sul_exp feed_lane_cleaning dust_mgmt no_classes
## no :33 scrape:40 no : 9 0 :38
## yes:86 lagoon:50 yes:110 1 :55
## clean :20 2 :26
## sl : 9 3 or more: 0
##
## # remove variables
mhaemall.analyze <- as.data.frame(mhaem.all)
mhaemall.analyze <- as.data.frame(mhaem.all) %>%
select(-"cef", -"gen", -"resp_vacc", -"colost_source", -"colost_past",
-"colost_again", -"hosp_milk_pastuerized", -"dust_mgmt") %>%
droplevels(mhaemall.analyze$no_classes, exclude = "3 or more")
# recategorize feed_lane_cleaning
mhaemall.analyze$feed_lane_cleaning <- factor(mhaem.raw$feed_lane_cleaning, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("scrape/lagoon", "scrape/lagoon",
"clean", "scrape/lagoon"))
# check
summary(mhaemall.analyze)## Farm season dan enr flr gam pen spc tet tild
## A:23 summer:60 S:36 S:34 S:79 S:75 S:74 S:100 S: 16 S:55
## B:26 winter:59 R:83 R:85 R:40 R:44 R:45 R: 19 R:103 R:64
## C:17
## D:20
## E: 9
## F:24
## tilm tul source_no onsite_milking salmonella first_def bovishield
## S:43 S:84 single :44 no :40 no :53 no :95 no :61
## R:76 R:35 low multi :49 yes:79 yes:66 yes:24 yes:58
## high multi:26
##
##
##
## pyramid titanium pinkeye clostrid action bull_calves milk_mass_med
## no :78 no :99 no :32 no :99 screen :102 no :52 no :63
## yes:41 yes:20 yes:87 yes:20 premium: 17 yes:67 yes:56
##
##
##
##
## mass_pen_exp mass_tet_exp mass_rum_exp mass_corr_exp mass_sul_exp
## no :96 no :24 no :23 no :67 no :33
## yes:23 yes:95 yes:96 yes:52 yes:86
##
##
##
##
## feed_lane_cleaning no_classes
## scrape/lagoon:99 0:38
## clean :20 1:55
## 2:26
##
##
## # create levels
mhaem.bl.all = tiers2blacklist(list("season", "Farm",
c("dan", "enr", "flr", "gam", "pen", "spc", "tet", "tild", "tilm",
"tul", "source_no", "onsite_milking", "first_def",
"salmonella","bovishield", "pyramid", "titanium", "pinkeye",
"clostrid", "action", "bull_calves",
"milk_mass_med", "mass_pen_exp", "mass_tet_exp", "mass_rum_exp",
"mass_corr_exp", "mass_sul_exp","feed_lane_cleaning",
"no_classes")))
mhaem.bl.all = rbind(mhaem.bl.all, c("season"))
mhaem.bl.all## from to
## [1,] "Farm" "season"
## [2,] "dan" "season"
## [3,] "enr" "season"
## [4,] "flr" "season"
## [5,] "gam" "season"
## [6,] "pen" "season"
## [7,] "spc" "season"
## [8,] "tet" "season"
## [9,] "tild" "season"
## [10,] "tilm" "season"
## [11,] "tul" "season"
## [12,] "source_no" "season"
## [13,] "onsite_milking" "season"
## [14,] "first_def" "season"
## [15,] "salmonella" "season"
## [16,] "bovishield" "season"
## [17,] "pyramid" "season"
## [18,] "titanium" "season"
## [19,] "pinkeye" "season"
## [20,] "clostrid" "season"
## [21,] "action" "season"
## [22,] "bull_calves" "season"
## [23,] "milk_mass_med" "season"
## [24,] "mass_pen_exp" "season"
## [25,] "mass_tet_exp" "season"
## [26,] "mass_rum_exp" "season"
## [27,] "mass_corr_exp" "season"
## [28,] "mass_sul_exp" "season"
## [29,] "feed_lane_cleaning" "season"
## [30,] "no_classes" "season"
## [31,] "dan" "Farm"
## [32,] "enr" "Farm"
## [33,] "flr" "Farm"
## [34,] "gam" "Farm"
## [35,] "pen" "Farm"
## [36,] "spc" "Farm"
## [37,] "tet" "Farm"
## [38,] "tild" "Farm"
## [39,] "tilm" "Farm"
## [40,] "tul" "Farm"
## [41,] "source_no" "Farm"
## [42,] "onsite_milking" "Farm"
## [43,] "first_def" "Farm"
## [44,] "salmonella" "Farm"
## [45,] "bovishield" "Farm"
## [46,] "pyramid" "Farm"
## [47,] "titanium" "Farm"
## [48,] "pinkeye" "Farm"
## [49,] "clostrid" "Farm"
## [50,] "action" "Farm"
## [51,] "bull_calves" "Farm"
## [52,] "milk_mass_med" "Farm"
## [53,] "mass_pen_exp" "Farm"
## [54,] "mass_tet_exp" "Farm"
## [55,] "mass_rum_exp" "Farm"
## [56,] "mass_corr_exp" "Farm"
## [57,] "mass_sul_exp" "Farm"
## [58,] "feed_lane_cleaning" "Farm"
## [59,] "no_classes" "Farm"
## [60,] "season" "season"2.2 BNA
# all farms hc structure learning
mhaemfarm.hc <- hc(mhaemall.analyze, blacklist = mhaem.bl.all, whitelist = wl, score = "aic")
# plot customization
mhaem.a <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemfarm.hc))
graph::nodeRenderInfo(mhaem.a) <- list(fontsize=60)
Rgraphviz::renderGraph(mhaem.a)
# all farms tabu structure learning
mhaemfarm.tabu <- tabu(mhaemall.analyze, blacklist = mhaem.bl.all,
whitelist = wl, score = "aic")
# plot customization
mhaem.b <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemfarm.tabu))
graph::nodeRenderInfo(mhaem.b) <- list(fontsize=60)
Rgraphviz::renderGraph(mhaem.b)
graphviz.compare(mhaemfarm.hc, mhaemfarm.tabu, shape = "ellipse",
main= c("Mhaem Farms HC DAG", "Mhaem Farms Tabu DAG"))
# bootstrap
mhaemfarm.boot = boot.strength(mhaemall.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = mhaem.bl.all,
whitelist = wl, score = "aic"))
# take arcs with strength of 50% and return average consensus network
mhaemfarm.avg = averaged.network(mhaemfarm.boot, threshold = 0.5)
# graph customization
mhaem.c <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemfarm.avg))
graph::nodeRenderInfo(mhaem.c) <- list(fontsize=60)
Rgraphviz::renderGraph(mhaem.c)
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(mhaemfarm.avg, mhaemfarm.hc, shape = "ellipse",
main= c("Mhaem Farm Level Factors Averaged DAG",
"Mhaem Farm Level Factors Single DAG"))
Animal level factors
2.3 Data Wrangling
# analytic data frame for individual risk factors
mhaem.ind <- mhaem.factor %>%
select("Farm", "season", "brd", "dan", "enr", "flr", "gam", "pen", "spc",
"tet", "tild", "tilm", "tul", "total_abx", "last_treat",
"days_since_pen", "days_since_ceph", "days_since_mac", "days_since_phen",
"days_since_tet", "days_since_sul", "days_since_flr")
# check for missingness
summary(aggr(mhaem.ind))
##
## Missings per variable:
## Variable Count
## Farm 0
## season 0
## brd 0
## dan 0
## enr 0
## flr 0
## gam 0
## pen 0
## spc 0
## tet 0
## tild 0
## tilm 0
## tul 0
## total_abx 9
## last_treat 9
## days_since_pen 9
## days_since_ceph 9
## days_since_mac 9
## days_since_phen 9
## days_since_tet 9
## days_since_sul 9
## days_since_flr 9
##
## Missings in combinations of variables:
## Combinations Count Percent
## 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0 110 92.436975
## 0:0:0:0:0:0:0:0:0:0:0:0:0:1:1:1:1:1:1:1:1:1 9 7.563025# check cell counts - All Farms
summary(mhaem.ind) # days_since_pen(pen yes/no), days_since_ceph(ceph yes/no), days_since_mac(mac yes/no), days_since_phen(0-60 days, >=60 days, no trx), days_since_tet(tet yes/no), days_since_sul(remove), days_since_flr(0-60 days, >= 60 days, no trx)## Farm season brd dan enr flr gam pen spc tet
## A:23 summer:60 neg:56 S:36 S:34 S:79 S:75 S:74 S:100 S: 16
## B:26 winter:59 pos:63 R:83 R:85 R:40 R:44 R:45 R: 19 R:103
## C:17
## D:20
## E: 9
## F:24
## tild tilm tul total_abx last_treat days_since_pen
## S:55 S:43 S:84 0 :30 <16 days :14 <16 days : 3
## R:64 R:76 R:35 1 to 2 :35 16-60 days:25 16-60 days: 4
## 3 or more:45 >=60 days :42 >=60 days :11
## NA's : 9 no trx :29 no trx :92
## NA's : 9 NA's : 9
##
## days_since_ceph days_since_mac days_since_phen days_since_tet
## <16 days : 1 <16 days : 2 <16 days : 8 16-60 days: 0
## 16-60 days: 2 16-60 days: 5 16-60 days:16 >=60 days :15
## >=60 days : 7 >=60 days :17 >=60 days :34 no trx :95
## no trx :100 no trx :86 no trx :52 NA's : 9
## NA's : 9 NA's : 9 NA's : 9
##
## days_since_sul days_since_flr
## >=60 days: 7 <16 days : 2
## no trx :103 16-60 days:11
## NA's : 9 >=60 days :16
## no trx :81
## NA's : 9
## # remove variables
mhaemind2 <- as.data.frame(mhaem.ind) %>%
select(-"days_since_sul")
# check
summary(mhaemind2)## Farm season brd dan enr flr gam pen spc tet
## A:23 summer:60 neg:56 S:36 S:34 S:79 S:75 S:74 S:100 S: 16
## B:26 winter:59 pos:63 R:83 R:85 R:40 R:44 R:45 R: 19 R:103
## C:17
## D:20
## E: 9
## F:24
## tild tilm tul total_abx last_treat days_since_pen
## S:55 S:43 S:84 0 :30 <16 days :14 <16 days : 3
## R:64 R:76 R:35 1 to 2 :35 16-60 days:25 16-60 days: 4
## 3 or more:45 >=60 days :42 >=60 days :11
## NA's : 9 no trx :29 no trx :92
## NA's : 9 NA's : 9
##
## days_since_ceph days_since_mac days_since_phen days_since_tet
## <16 days : 1 <16 days : 2 <16 days : 8 16-60 days: 0
## 16-60 days: 2 16-60 days: 5 16-60 days:16 >=60 days :15
## >=60 days : 7 >=60 days :17 >=60 days :34 no trx :95
## no trx :100 no trx :86 no trx :52 NA's : 9
## NA's : 9 NA's : 9 NA's : 9
##
## days_since_flr
## <16 days : 2
## 16-60 days:11
## >=60 days :16
## no trx :81
## NA's : 9
## # recategorize
mhaemind2$days_since_pen <- factor(mhaem.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
mhaemind2$days_since_ceph <- factor(mhaem.raw$days_since_ceph, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
mhaemind2$days_since_mac <- factor(mhaem.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
mhaemind2$days_since_phen <- factor(mhaem.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("0-60 days", "0-60 days",
">=60 days", "no treatment"))
mhaemind2$days_since_tet <- factor(mhaem.raw$days_since_tet, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
mhaemind2$days_since_flr <- factor(mhaem.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("0-60 days", "0-60 days",
">=60 days", "no treatment"))
# remove Farm E
mhaemind2$Farm <- factor(mhaem.raw$Farm, order = FALSE,
levels = c(1, 2, 3, 4, 6),
labels = c("A", "B", "C", "D", "F"))
# summary(mhaemind2)
mhaemind.analyze <- na.omit(mhaemind2)
# check
summary(mhaemind.analyze)## Farm season brd dan enr flr gam pen spc tet
## A:23 summer:55 neg:50 S:35 S:34 S:72 S:71 S:70 S:93 S:15
## B:26 winter:55 pos:60 R:75 R:76 R:38 R:39 R:40 R:17 R:95
## C:17
## D:20
## F:24
## tild tilm tul total_abx last_treat days_since_pen
## S:55 S:42 S:78 0 :30 <16 days :14 treated :18
## R:55 R:68 R:32 1 to 2 :35 16-60 days:25 no treatment:92
## 3 or more:45 >=60 days :42
## no trx :29
##
## days_since_ceph days_since_mac days_since_phen
## treated : 10 treated :24 0-60 days :24
## no treatment:100 no treatment:86 >=60 days :34
## no treatment:52
##
##
## days_since_tet days_since_flr
## treated :15 0-60 days :13
## no treatment:95 >=60 days :16
## no treatment:81
##
## # create levels
mhaem.bl.ind = tiers2blacklist(list("season", "Farm",
c("brd", "dan", "enr", "flr", "gam", "pen", "spc", "tet", "tild", "tilm",
"tul", "total_abx", "last_treat", "days_since_pen",
"days_since_ceph", "days_since_mac", "days_since_phen",
"days_since_tet", "days_since_flr")))
mhaem.bl.ind = rbind(mhaem.bl.ind, c("season"))
mhaem.bl.ind## from to
## [1,] "Farm" "season"
## [2,] "brd" "season"
## [3,] "dan" "season"
## [4,] "enr" "season"
## [5,] "flr" "season"
## [6,] "gam" "season"
## [7,] "pen" "season"
## [8,] "spc" "season"
## [9,] "tet" "season"
## [10,] "tild" "season"
## [11,] "tilm" "season"
## [12,] "tul" "season"
## [13,] "total_abx" "season"
## [14,] "last_treat" "season"
## [15,] "days_since_pen" "season"
## [16,] "days_since_ceph" "season"
## [17,] "days_since_mac" "season"
## [18,] "days_since_phen" "season"
## [19,] "days_since_tet" "season"
## [20,] "days_since_flr" "season"
## [21,] "brd" "Farm"
## [22,] "dan" "Farm"
## [23,] "enr" "Farm"
## [24,] "flr" "Farm"
## [25,] "gam" "Farm"
## [26,] "pen" "Farm"
## [27,] "spc" "Farm"
## [28,] "tet" "Farm"
## [29,] "tild" "Farm"
## [30,] "tilm" "Farm"
## [31,] "tul" "Farm"
## [32,] "total_abx" "Farm"
## [33,] "last_treat" "Farm"
## [34,] "days_since_pen" "Farm"
## [35,] "days_since_ceph" "Farm"
## [36,] "days_since_mac" "Farm"
## [37,] "days_since_phen" "Farm"
## [38,] "days_since_tet" "Farm"
## [39,] "days_since_flr" "Farm"
## [40,] "season" "season"2.4 BNA
# individual factors hc structure learning
mhaemind.hc <- hc(mhaemind.analyze, blacklist = mhaem.bl.ind,
whitelist = wl, score = "aic")
# plot customization
mhaem.c <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemind.hc))
graph::nodeRenderInfo(mhaem.c) <- list(fontsize=60)
Rgraphviz::renderGraph(mhaem.c)
# individual factors tabu structure learning
mhaemind.tabu <- tabu(mhaemind.analyze, blacklist = mhaem.bl.ind,
whitelist = wl, score = "aic")
# plot customization
mhaem.d <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemind.tabu))
graph::nodeRenderInfo(mhaem.d) <- list(fontsize=60)
Rgraphviz::renderGraph(mhaem.d)
graphviz.compare(mhaemind.hc, mhaemind.tabu, shape = "ellipse",
main= c("Mhaem Individual Risk Factors HC DAG",
"Mhaem Individual Risk Factors Tabu DAG"))
# bootstrap
mhaemind.boot = boot.strength(mhaemind.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = mhaem.bl.ind,
whitelist = wl, score = "aic"))
# take arcs with strength of 50% and return average consensus network
mhaemind.avg = averaged.network(mhaemind.boot, threshold = 0.5)
# graph customization
mhaem.e <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(mhaemind.avg))
graph::nodeRenderInfo(mhaem.e) <- list(fontsize=40)
Rgraphviz::renderGraph(mhaem.e)
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(mhaemind.avg, mhaemind.hc, shape = "ellipse",
main= c("M haem Individual Level Factors Averaged DAG",
"M haem Individual Level Factors Single DAG"))
2.5 Parameter Learning
Farm level query
# fit the parameters - bayesian method
mhaemfarm.bayes = bn.fit(mhaemfarm.avg, mhaemall.analyze, method = "bayes")
# probability table
mhaemfarm.bayes##
## Bayesian network parameters
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.10055096 0.28711485
## B 0.26584022 0.16946779
## C 0.13360882 0.15266106
## D 0.21625344 0.11904762
## E 0.08402204 0.06862745
## F 0.19972452 0.20308123
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.5041667 0.4958333
##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## dan A B C D E F
## S 0.089928058 0.079617834 0.179611650 0.995867769 0.118181818 0.334482759
## R 0.910071942 0.920382166 0.820388350 0.004132231 0.881818182 0.665517241
##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## , , pinkeye = no
##
## dan
## enr S R
## S 0.346153846 0.004273504
## R 0.653846154 0.995726496
##
## , , pinkeye = yes
##
## dan
## enr S R
## S 0.996240602 0.002304147
## R 0.003759398 0.997695853
##
##
## Parameters of node flr (multinomial distribution)
##
## Conditional probability table:
##
## , , pen = S, tul = S
##
## Farm
## flr A B C D E F
## S 0.998960499 0.998402556 0.797520661 0.980000000 0.993150685 0.998113208
## R 0.001039501 0.001597444 0.202479339 0.020000000 0.006849315 0.001886792
##
## , , pen = R, tul = S
##
## Farm
## flr A B C D E F
## S 0.993150685 0.989795918 0.989795918 0.263676149 0.993150685 0.989795918
## R 0.006849315 0.010204082 0.010204082 0.736323851 0.006849315 0.010204082
##
## , , pen = S, tul = R
##
## Farm
## flr A B C D E F
## S 0.500000000 0.334482759 0.335616438 0.500000000 0.980000000 0.545283019
## R 0.500000000 0.665517241 0.664383562 0.500000000 0.020000000 0.454716981
##
## , , pen = R, tul = R
##
## Farm
## flr A B C D E F
## S 0.500000000 0.004132231 0.002958580 0.500000000 0.010204082 0.500000000
## R 0.500000000 0.995867769 0.997041420 0.500000000 0.989795918 0.500000000
##
##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D E F
## S 0.996402878 0.576433121 0.354368932 0.995867769 0.445454545 0.293103448
## R 0.003597122 0.423566879 0.645631068 0.004132231 0.554545455 0.706896552
##
## Parameters of node pen (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pen A B C D E F
## S 0.86690647 0.72929936 0.47087379 0.05371901 0.44545455 0.91379310
## R 0.13309353 0.27070064 0.52912621 0.94628099 0.55454545 0.08620690
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## tilm
## spc S R
## S 0.994252874 0.748366013
## R 0.005747126 0.251633987
##
## Parameters of node tet (multinomial distribution)
##
## Conditional probability table:
##
## enr
## tet S R
## S 0.44202899 0.01461988
## R 0.55797101 0.98538012
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , gam = S
##
## enr
## tild S R
## S 0.996240602 0.500000000
## R 0.003759398 0.500000000
##
## , , gam = R
##
## enr
## tild S R
## S 0.100000000 0.026011561
## R 0.900000000 0.973988439
##
##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## , , tild = S
##
## pen
## tilm S R
## S 0.932000000 0.541237113
## R 0.068000000 0.458762887
##
## , , tild = R
##
## pen
## tilm S R
## S 0.002890173 0.052941176
## R 0.997109827 0.947058824
##
##
## Parameters of node tul (multinomial distribution)
##
## Conditional probability table:
##
## gam
## tul S R
## S 0.98344371 0.23033708
## R 0.01655629 0.76966292
##
## Parameters of node source_no (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## source_no A B C D E
## single 0.002398082 0.002123142 0.003236246 0.994490358 0.006060606
## low multi 0.995203837 0.995753715 0.003236246 0.002754821 0.006060606
## high multi 0.002398082 0.002123142 0.993527508 0.002754821 0.987878788
## Farm
## source_no F
## single 0.995402299
## low multi 0.002298851
## high multi 0.002298851
##
## Parameters of node onsite_milking (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## onsite_milking A B C D E
## no 0.996402878 0.003184713 0.995145631 0.004132231 0.009090909
## yes 0.003597122 0.996815287 0.004854369 0.995867769 0.990909091
## Farm
## onsite_milking F
## no 0.003448276
## yes 0.996551724
##
## Parameters of node salmonella (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## salmonella A B C D E
## no 0.003597122 0.003184713 0.004854369 0.995867769 0.990909091
## yes 0.996402878 0.996815287 0.995145631 0.004132231 0.009090909
## Farm
## salmonella F
## no 0.996551724
## yes 0.003448276
##
## Parameters of node first_def (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## first_def A B C D E
## no 0.996402878 0.996815287 0.995145631 0.995867769 0.990909091
## yes 0.003597122 0.003184713 0.004854369 0.004132231 0.009090909
## Farm
## first_def F
## no 0.003448276
## yes 0.996551724
##
## Parameters of node bovishield (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bovishield A B C D E
## no 0.003597122 0.003184713 0.995145631 0.995867769 0.009090909
## yes 0.996402878 0.996815287 0.004854369 0.004132231 0.990909091
## Farm
## bovishield F
## no 0.996551724
## yes 0.003448276
##
## Parameters of node pyramid (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pyramid A B C D E F
## no 0.996402878 0.996815287 0.004854369 0.995867769 0.990909091 0.003448276
## yes 0.003597122 0.003184713 0.995145631 0.004132231 0.009090909 0.996551724
##
## Parameters of node titanium (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## titanium A B C D E
## no 0.996402878 0.996815287 0.995145631 0.004132231 0.990909091
## yes 0.003597122 0.003184713 0.004854369 0.995867769 0.009090909
## Farm
## titanium F
## no 0.996551724
## yes 0.003448276
##
## Parameters of node pinkeye (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pinkeye A B C D E F
## no 0.996402878 0.003184713 0.004854369 0.004132231 0.990909091 0.003448276
## yes 0.003597122 0.996815287 0.995145631 0.995867769 0.009090909 0.996551724
##
## Parameters of node clostrid (multinomial distribution)
##
## Conditional probability table:
##
## titanium
## clostrid no yes
## no 0.997487437 0.012195122
## yes 0.002512563 0.987804878
##
## Parameters of node action (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## action A B C D E
## screen 0.996402878 0.996815287 0.004854369 0.995867769 0.990909091
## premium 0.003597122 0.003184713 0.995145631 0.004132231 0.009090909
## Farm
## action F
## screen 0.996551724
## premium 0.003448276
##
## Parameters of node bull_calves (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bull_calves A B C D E
## no 0.003597122 0.996815287 0.995145631 0.004132231 0.990909091
## yes 0.996402878 0.003184713 0.004854369 0.995867769 0.009090909
## Farm
## bull_calves F
## no 0.003448276
## yes 0.996551724
##
## Parameters of node milk_mass_med (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## milk_mass_med A B C D E
## no 0.003597122 0.996815287 0.995145631 0.995867769 0.009090909
## yes 0.996402878 0.003184713 0.004854369 0.004132231 0.990909091
## Farm
## milk_mass_med F
## no 0.003448276
## yes 0.996551724
##
## Parameters of node mass_pen_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_pen_exp A B C D E
## no 0.003597122 0.996815287 0.995145631 0.995867769 0.990909091
## yes 0.996402878 0.003184713 0.004854369 0.004132231 0.009090909
## Farm
## mass_pen_exp F
## no 0.996551724
## yes 0.003448276
##
## Parameters of node mass_tet_exp (multinomial distribution)
##
## Conditional probability table:
##
## first_def
## mass_tet_exp no yes
## no 0.002617801 0.989795918
## yes 0.997382199 0.010204082
##
## Parameters of node mass_rum_exp (multinomial distribution)
##
## Conditional probability table:
##
## mass_pen_exp
## mass_rum_exp no yes
## no 0.002590674 0.989361702
## yes 0.997409326 0.010638298
##
## Parameters of node mass_corr_exp (multinomial distribution)
##
## Conditional probability table:
##
## bull_calves
## mass_corr_exp no yes
## no 0.004761905 0.996296296
## yes 0.995238095 0.003703704
##
## Parameters of node mass_sul_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_sul_exp A B C D E
## no 0.003597122 0.003184713 0.004854369 0.004132231 0.990909091
## yes 0.996402878 0.996815287 0.995145631 0.995867769 0.009090909
## Farm
## mass_sul_exp F
## no 0.996551724
## yes 0.003448276
##
## Parameters of node feed_lane_cleaning (multinomial distribution)
##
## Conditional probability table:
##
## titanium
## feed_lane_cleaning no yes
## scrape/lagoon 0.997487437 0.012195122
## clean 0.002512563 0.987804878
##
## Parameters of node no_classes (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## no_classes A B C D E
## 0 0.606714628 0.116772824 0.119741100 0.101928375 0.987878788
## 1 0.390887290 0.460721868 0.702265372 0.498622590 0.006060606
## 2 0.002398082 0.422505308 0.177993528 0.399449036 0.006060606
## Farm
## no_classes F
## 0 0.333333333
## 1 0.498850575
## 2 0.167816092set.seed(117)
cpquery(mhaemfarm.bayes, (enr == "R"), pinkeye == "yes")## [1] 0.622962cpquery(mhaemfarm.bayes, (enr == "R"), pinkeye == "no")## [1] 0.962188cpquery(mhaemfarm.bayes, (enr == "S"), pinkeye == "yes")## [1] 0.3785272cpquery(mhaemfarm.bayes, (enr == "S"), pinkeye == "no")## [1] 0.04271263Animal level query
# fit the parameters - bayesian method
mhaemind.bayes = bn.fit(mhaemind.avg, mhaemind.analyze, method = "bayes")
# probability table
mhaemind.bayes##
## Bayesian network parameters
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.1099099 0.3081081
## B 0.2900901 0.1819820
## C 0.1459459 0.1639640
## D 0.2360360 0.1279279
## F 0.2180180 0.2180180
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.5 0.5
##
## Parameters of node brd (multinomial distribution)
##
## Conditional probability table:
## neg pos
## 0.454955 0.545045
##
## Parameters of node dan (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## dan A B C D F
## S 0.090517241 0.080152672 0.180232558 0.995049505 0.334710744
## R 0.909482759 0.919847328 0.819767442 0.004950495 0.665289256
##
## Parameters of node enr (multinomial distribution)
##
## Conditional probability table:
##
## , , tild = S
##
## dan
## enr S R
## S 0.996240602 0.005617978
## R 0.003759398 0.994382022
##
## , , tild = R
##
## dan
## enr S R
## S 0.500000000 0.002347418
## R 0.500000000 0.997652582
##
##
## Parameters of node flr (multinomial distribution)
##
## Conditional probability table:
##
## , , tul = S
##
## Farm
## flr A B C D F
## S 0.997835498 0.996688742 0.852112676 0.300995025 0.996183206
## R 0.002164502 0.003311258 0.147887324 0.699004975 0.003816794
##
## , , tul = R
##
## Farm
## flr A B C D F
## S 0.500000000 0.184684685 0.103960396 0.500000000 0.545045045
## R 0.500000000 0.815315315 0.896039604 0.500000000 0.454954955
##
##
## Parameters of node gam (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## gam A B C D F
## S 0.995689655 0.576335878 0.354651163 0.995049505 0.293388430
## R 0.004310345 0.423664122 0.645348837 0.004950495 0.706611570
##
## Parameters of node pen (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pen A B C D F
## S 0.86637931 0.72900763 0.47093023 0.05445545 0.91322314
## R 0.13362069 0.27099237 0.52906977 0.94554455 0.08677686
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## spc A B C D F
## S 0.909482759 0.576335878 0.761627907 0.995049505 0.995867769
## R 0.090517241 0.423664122 0.238372093 0.004950495 0.004132231
##
## Parameters of node tet (multinomial distribution)
##
## Conditional probability table:
##
## , , flr = S
##
## enr
## tet S R
## S 0.697530864 0.002392344
## R 0.302469136 0.997607656
##
## , , flr = R
##
## enr
## tet S R
## S 0.078947368 0.005154639
## R 0.921052632 0.994845361
##
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## gam
## tild S R
## S 0.75874126 0.03164557
## R 0.24125874 0.96835443
##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## , , tild = S
##
## pen
## tilm S R
## S 0.932000000 0.541237113
## R 0.068000000 0.458762887
##
## , , tild = R
##
## pen
## tilm S R
## S 0.003184713 0.007692308
## R 0.996815287 0.992307692
##
##
## Parameters of node tul (multinomial distribution)
##
## Conditional probability table:
##
## gam
## tul S R
## S 0.996503497 0.183544304
## R 0.003496503 0.816455696
##
## Parameters of node total_abx (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## total_abx A B C D F
## 0 0.64942529 0.11704835 0.12015504 0.10231023 0.33333333
## 1 to 2 0.30459770 0.23155216 0.46899225 0.44884488 0.20936639
## 3 or more 0.04597701 0.65139949 0.41085271 0.44884488 0.45730028
##
## Parameters of node last_treat (multinomial distribution)
##
## Conditional probability table:
##
## total_abx
## last_treat 0 1 to 2 3 or more
## <16 days 0.002747253 0.172169811 0.178308824
## 16-60 days 0.002747253 0.172169811 0.420955882
## >=60 days 0.035714286 0.653301887 0.398897059
## no trx 0.958791209 0.002358491 0.001838235
##
## Parameters of node days_since_pen (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_pen A B C D F
## treated 0.004310345 0.423664122 0.412790698 0.004950495 0.004132231
## no treatment 0.995689655 0.576335878 0.587209302 0.995049505 0.995867769
##
## Parameters of node days_since_ceph (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_ceph A B C D F
## treated 0.047413793 0.003816794 0.005813953 0.054455446 0.334710744
## no treatment 0.952586207 0.996183206 0.994186047 0.945544554 0.665289256
##
## Parameters of node days_since_mac (multinomial distribution)
##
## Conditional probability table:
##
## total_abx
## days_since_mac 0 1 to 2 3 or more
## treated 0.005494505 0.089622642 0.466911765
## no treatment 0.994505495 0.910377358 0.533088235
##
## Parameters of node days_since_phen (multinomial distribution)
##
## Conditional probability table:
##
## , , days_since_pen = treated
##
## last_treat
## days_since_phen <16 days 16-60 days >=60 days no trx
## 0-60 days 0.788617886 0.695473251 0.013333333 0.333333333
## >=60 days 0.203252033 0.004115226 0.013333333 0.333333333
## no treatment 0.008130081 0.300411523 0.973333333 0.333333333
##
## , , days_since_pen = no treatment
##
## last_treat
## days_since_phen <16 days 16-60 days >=60 days no trx
## 0-60 days 0.771689498 0.399449036 0.001064963 0.001430615
## >=60 days 0.114155251 0.465564738 0.640042599 0.001430615
## no treatment 0.114155251 0.134986226 0.358892439 0.997138770
##
##
## Parameters of node days_since_tet (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## days_since_tet A B C D F
## treated 0.004310345 0.003816794 0.005813953 0.648514851 0.086776860
## no treatment 0.995689655 0.996183206 0.994186047 0.351485149 0.913223140
##
## Parameters of node days_since_flr (multinomial distribution)
##
## Conditional probability table:
##
## last_treat
## days_since_flr <16 days 16-60 days >=60 days no trx
## 0-60 days 0.356725146 0.320132013 0.001972387 0.002849003
## >=60 days 0.076023392 0.161716172 0.262327416 0.002849003
## no treatment 0.567251462 0.518151815 0.735700197 0.9943019943. H. somni
Farm level factors
3.1 Data Wrangling
# import data
hsom.df <- read_csv("HSOM.csv", show_col_types = FALSE)
# remove na rows
hsom.df <- filter(hsom.df, rowSums(is.na(hsom.df)) != ncol(hsom.df))# connect calf treatment variables to hsom df
hsom.df <- hsom.df %>%
mutate(calf_ID = ID) %>%
select(-"ID")
hsom.raw <- left_join(hsom.df, calf.trx, by = "calf_ID")
# make variables factors, add labels
hsom.factor = hsom.raw
hsom.factor$season <- factor(hsom.raw$season, order = FALSE,
levels = c(1, 2),
labels = c("summer", "winter"))
hsom.factor$brd <- factor(hsom.raw$brd, order = FALSE,
levels = c(0, 1),
labels = c("neg", "pos"))
hsom.factor$cef <- factor(hsom.raw$cef, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$dan <- factor(hsom.raw$dan, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$enr <- factor(hsom.raw$enr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$flr <- factor(hsom.raw$flr, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$gam <- factor(hsom.raw$gam, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$gen <- factor(hsom.raw$gen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$pen <- factor(hsom.raw$pen, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$spc <- factor(hsom.raw$spc, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$tet <- factor(hsom.raw$tet, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$tild <- factor(hsom.raw$tild, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$tilm <- factor(hsom.raw$tilm, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$tul <- factor(hsom.raw$tul, order = FALSE,
levels = c(0, 1),
labels = c("S", "R"))
hsom.factor$source_no <- factor(hsom.raw$source_no, order = FALSE,
levels = c(1, 2, 3),
labels = c("single", "low multi", "high multi"))
hsom.factor$onsite_milking <- factor(hsom.raw$onsite_milking, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$resp_vacc <- factor(hsom.raw$resp_vacc, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$salmonella <- factor(hsom.raw$salmonella, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$bovishield <- factor(hsom.raw$bovishield, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$pyramid <- factor(hsom.raw$pyramid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$titanium <- factor(hsom.raw$titanium, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$pinkeye <- factor(hsom.raw$pinkeye, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$clostrid <- factor(hsom.raw$clostrid, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$colost_source <- factor(hsom.raw$colost_source, order = FALSE,
levels = c(1, 2),
labels = c("p", "dp"))
hsom.factor$colost_past <- factor(hsom.raw$colost_past, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$colost_again <- factor(hsom.raw$colost_again, order = FALSE,
levels = c(0, 1, 99),
labels = c("no", "yes", "-99"))
hsom.factor$hosp_milk_pastuerized <- factor(hsom.raw$hosp_milk_pastuerized,
order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$action <- factor(hsom.raw$action, order = FALSE,
levels = c(1, 2),
labels = c("screen", "premium"))
hsom.factor$bull_calves <- factor(hsom.raw$bull_calves, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$milk_mass_med <- factor(hsom.raw$milk_mass_med, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mass_pen_exp <- factor(hsom.raw$mass_pen_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mass_tet_exp <- factor(hsom.raw$mass_tet_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mass_corr_exp <- factor(hsom.raw$mass_corr_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mass_rum_exp <- factor(hsom.raw$mass_rum_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mass_sul_exp <- factor(hsom.raw$mass_sul_exp, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$feed_lane_cleaning <- factor(hsom.raw$feed_lane_cleaning, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("scrape", "lagoon", "clean", "sl"))
hsom.factor$dust_mgmt <- factor(hsom.raw$dust_mgmt, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$mort_rate <- factor(hsom.raw$mort_rate, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$pen_trx_rate <- factor(hsom.raw$pen_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$cef_trx_rate <- factor(hsom.raw$cef_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$mac_trx_rate <- factor(hsom.raw$mac_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$res_nuf_trx_rate <- factor(hsom.raw$res_nuf_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$tet_inj_use <- factor(hsom.raw$tet_inj_use, order = FALSE,
levels = c(0, 1),
labels = c("no", "yes"))
hsom.factor$sul_trx_rate <- factor(hsom.raw$sul_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$bay_trx_rate <- factor(hsom.raw$bay_trx_rate, order = FALSE,
levels = c(1, 2),
labels = c("low", "high"))
hsom.factor$farm <- factor(hsom.raw$farm, order = FALSE,
levels = c(1, 2, 3, 4, 5, 6),
labels = c("A", "B", "C", "D", "E", "F"))
hsom.factor$last_treat <- factor(hsom.raw$last_treat, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days", "no trx"))
hsom.factor$no_classes <- factor(hsom.raw$no_classes, order = FALSE,
levels = c(0, 1, 2),
labels = c("0", "1", "2 or more"))
hsom.factor$total_abx <- factor(hsom.raw$total_abx, order = FALSE,
levels = c(0, 1, 2),
labels = c("0", "1 to 2", "3 or more"))
hsom.factor$days_since_pen <- factor(hsom.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
hsom.factor$days_since_ceph <- factor(hsom.raw$days_since_ceph, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
hsom.factor$days_since_mac <- factor(hsom.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
hsom.factor$days_since_phen <- factor(hsom.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<16 days", "16-60 days", ">=60 days",
"no trx"))
hsom.factor$days_since_tet <- factor(hsom.raw$days_since_tet, order = FALSE,
levels = c(2, 3, 4),
labels = c("16-60 days", ">=60 days", "no trx"))
hsom.factor$days_since_sul <- factor(hsom.raw$days_since_sul, order = FALSE,
levels = c(3, 4),
labels = c(">=60 days", "no trx"))
hsom.factor$days_since_flr <- factor(hsom.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated", ">=60 days", "no trx"))
# glimpse(hsom.factor)# analytic data frame for all farms
hsom.all <- hsom.factor %>%
mutate(Farm = farm) %>%
select(-"farm", -"srp", -"entervene", -"pen_trx_rate", -"cef_trx_rate",
-"mac_trx_rate", -"res_nuf_trx_rate", -"brd",
-"tet_inj_use", -"sul_trx_rate", -"bay_trx_rate", -"mort_rate",
- "calf_ID", -"Farm_ID", -"cef", -"dan", -"enr", -"flr", -"total_abx",
-"last_treat", -"days_since_pen", -"days_since_ceph", -"colost_again",
-"days_since_mac", -"days_since_phen", -"days_since_tet",
-"days_since_sul", -"days_since_flr", -"gen")
# check for missingness
summary(aggr(hsom.all))
##
## Missings per variable:
## Variable Count
## season 0
## gam 0
## pen 0
## spc 0
## tet 0
## tild 0
## tilm 0
## tul 0
## source_no 0
## onsite_milking 0
## resp_vacc 0
## salmonella 0
## bovishield 0
## pyramid 0
## titanium 0
## pinkeye 0
## clostrid 0
## colost_source 0
## colost_past 0
## hosp_milk_pastuerized 0
## action 0
## bull_calves 0
## milk_mass_med 0
## mass_pen_exp 0
## mass_tet_exp 0
## mass_rum_exp 0
## mass_corr_exp 0
## mass_sul_exp 0
## feed_lane_cleaning 0
## dust_mgmt 0
## no_classes 0
## Farm 0
##
## Missings in combinations of variables:
## Combinations Count Percent
## 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0 97 100# check cell counts - All Farms
summary(hsom.all) # gam, pen, tul, mass_tet_exp, no_classes(3 or more) below 10## season gam pen spc tet tild tilm tul source_no
## summer:43 S:93 S:94 S:29 S:36 S:82 S:86 S:93 single :23
## winter:54 R: 4 R: 3 R:68 R:61 R:15 R:11 R: 4 low multi :22
## high multi:52
##
##
##
## onsite_milking resp_vacc salmonella bovishield pyramid titanium pinkeye
## no :31 no :32 no :54 no :43 no :69 no :82 no :43
## yes:66 yes:65 yes:43 yes:54 yes:28 yes:15 yes:54
##
##
##
##
## clostrid colost_source colost_past hosp_milk_pastuerized action
## no :82 p :64 no :32 no :32 screen :77
## yes:15 dp:33 yes:65 yes:65 premium:20
##
##
##
##
## bull_calves milk_mass_med mass_pen_exp mass_tet_exp mass_rum_exp mass_corr_exp
## no :63 no :46 no :86 no : 8 no :11 no :34
## yes:34 yes:51 yes:11 yes:89 yes:86 yes:63
##
##
##
##
## mass_sul_exp feed_lane_cleaning dust_mgmt no_classes Farm
## no :40 scrape:31 no :32 0 :41 A:11
## yes:57 lagoon:19 yes:65 1 :35 B:11
## clean :15 2 or more:21 C:20
## sl :32 D:15
## E:32
## F: 8# remove variables
hsomall.analyze <- as.data.frame(hsom.all)
hsomall.analyze <- as.data.frame(hsom.all) %>%
select(-"gam", -"pen", - "tul",
-"mass_tet_exp") %>%
droplevels(hsomall.analyze$no_classes, exclude = "3 or more")
# check
summary(hsomall.analyze)## season spc tet tild tilm source_no onsite_milking
## summer:43 S:29 S:36 S:82 S:86 single :23 no :31
## winter:54 R:68 R:61 R:15 R:11 low multi :22 yes:66
## high multi:52
##
##
##
## resp_vacc salmonella bovishield pyramid titanium pinkeye clostrid
## no :32 no :54 no :43 no :69 no :82 no :43 no :82
## yes:65 yes:43 yes:54 yes:28 yes:15 yes:54 yes:15
##
##
##
##
## colost_source colost_past hosp_milk_pastuerized action bull_calves
## p :64 no :32 no :32 screen :77 no :63
## dp:33 yes:65 yes:65 premium:20 yes:34
##
##
##
##
## milk_mass_med mass_pen_exp mass_rum_exp mass_corr_exp mass_sul_exp
## no :46 no :86 no :11 no :34 no :40
## yes:51 yes:11 yes:86 yes:63 yes:57
##
##
##
##
## feed_lane_cleaning dust_mgmt no_classes Farm
## scrape:31 no :32 0 :41 A:11
## lagoon:19 yes:65 1 :35 B:11
## clean :15 2 or more:21 C:20
## sl :32 D:15
## E:32
## F: 8# create levels
hsom.bl.all = tiers2blacklist(list("season", "Farm",
c("spc", "tet", "tild", "tilm",
"source_no", "onsite_milking", "resp_vacc",
"salmonella", "bovishield", "pyramid", "titanium", "pinkeye",
"clostrid", "colost_source", "colost_past",
"hosp_milk_pastuerized", "action", "bull_calves", "milk_mass_med",
"mass_pen_exp", "mass_rum_exp", "mass_corr_exp", "mass_sul_exp",
"feed_lane_cleaning", "dust_mgmt", "no_classes")))
hsom.bl.all = rbind(hsom.bl.all, c("season"))
hsom.bl.all## from to
## [1,] "Farm" "season"
## [2,] "spc" "season"
## [3,] "tet" "season"
## [4,] "tild" "season"
## [5,] "tilm" "season"
## [6,] "source_no" "season"
## [7,] "onsite_milking" "season"
## [8,] "resp_vacc" "season"
## [9,] "salmonella" "season"
## [10,] "bovishield" "season"
## [11,] "pyramid" "season"
## [12,] "titanium" "season"
## [13,] "pinkeye" "season"
## [14,] "clostrid" "season"
## [15,] "colost_source" "season"
## [16,] "colost_past" "season"
## [17,] "hosp_milk_pastuerized" "season"
## [18,] "action" "season"
## [19,] "bull_calves" "season"
## [20,] "milk_mass_med" "season"
## [21,] "mass_pen_exp" "season"
## [22,] "mass_rum_exp" "season"
## [23,] "mass_corr_exp" "season"
## [24,] "mass_sul_exp" "season"
## [25,] "feed_lane_cleaning" "season"
## [26,] "dust_mgmt" "season"
## [27,] "no_classes" "season"
## [28,] "spc" "Farm"
## [29,] "tet" "Farm"
## [30,] "tild" "Farm"
## [31,] "tilm" "Farm"
## [32,] "source_no" "Farm"
## [33,] "onsite_milking" "Farm"
## [34,] "resp_vacc" "Farm"
## [35,] "salmonella" "Farm"
## [36,] "bovishield" "Farm"
## [37,] "pyramid" "Farm"
## [38,] "titanium" "Farm"
## [39,] "pinkeye" "Farm"
## [40,] "clostrid" "Farm"
## [41,] "colost_source" "Farm"
## [42,] "colost_past" "Farm"
## [43,] "hosp_milk_pastuerized" "Farm"
## [44,] "action" "Farm"
## [45,] "bull_calves" "Farm"
## [46,] "milk_mass_med" "Farm"
## [47,] "mass_pen_exp" "Farm"
## [48,] "mass_rum_exp" "Farm"
## [49,] "mass_corr_exp" "Farm"
## [50,] "mass_sul_exp" "Farm"
## [51,] "feed_lane_cleaning" "Farm"
## [52,] "dust_mgmt" "Farm"
## [53,] "no_classes" "Farm"
## [54,] "season" "season"3.2 BNA
# all farms hc structure learning
hsomfarm.hc <- hc(hsomall.analyze, blacklist = hsom.bl.all,
whitelist = wl, score = "aic")
# plot customization
hsom.a <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomfarm.hc))
graph::nodeRenderInfo(hsom.a) <- list(fontsize=50)
Rgraphviz::renderGraph(hsom.a)
# all farms tabu structure learning
hsomfarm.tabu <- tabu(hsomall.analyze, blacklist = hsom.bl.all, whitelist = wl, score = "aic")
# plot customization
hsom.b <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomfarm.tabu))
graph::nodeRenderInfo(hsom.b) <- list(fontsize=50)
Rgraphviz::renderGraph(hsom.b)
graphviz.compare(hsomfarm.hc, hsomfarm.tabu, shape = "ellipse",
main= c("Hsom Farms HC DAG", "Hsom Farms Tabu DAG"))
# bootstrap
hsomfarm.boot = boot.strength(hsomall.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = hsom.bl.all,
whitelist = wl, score = "aic"))
# take arcs with strength of 50% and return average consensus network
hsomfarm.avg = averaged.network(hsomfarm.boot, threshold = 0.5)
# graph customization
hsom.c <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomfarm.avg))
graph::nodeRenderInfo(hsom.c) <- list(fontsize=50)
Rgraphviz::renderGraph(hsom.c)
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(hsomfarm.avg, hsomfarm.hc, shape = "ellipse",
main= c("Hsom Farm Level Factors Averaged DAG",
"Hsom Farm Level Factors Single DAG"))
Animal level factors
3.3 Data Wrangling
# analytic data frame for individual risk factors
hsom.ind <- hsom.factor %>%
mutate(Farm = farm) %>%
select("Farm", "season", "brd", "dan", "enr", "flr", "gam", "pen", "spc",
"tet", "tild", "tilm", "tul", "total_abx", "last_treat",
"days_since_pen", "days_since_ceph", "days_since_mac", "days_since_phen",
"days_since_tet", "days_since_sul", "days_since_flr")
# check for missingness
summary(aggr(hsom.ind))
##
## Missings per variable:
## Variable Count
## Farm 0
## season 0
## brd 0
## dan 0
## enr 0
## flr 0
## gam 0
## pen 0
## spc 0
## tet 0
## tild 0
## tilm 0
## tul 0
## total_abx 32
## last_treat 32
## days_since_pen 32
## days_since_ceph 32
## days_since_mac 32
## days_since_phen 32
## days_since_tet 32
## days_since_sul 32
## days_since_flr 32
##
## Missings in combinations of variables:
## Combinations Count Percent
## 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0 65 67.01031
## 0:0:0:0:0:0:0:0:0:0:0:0:0:1:1:1:1:1:1:1:1:1 32 32.98969# check cell counts - All Farms
summary(hsom.ind) # -dan, -enr, -flr, -gam, -pen, -tul, total_abx(<=2, 3 or more), last_treat(0-60 days, >=60 days, none), days_since_pen(pen yes/no), -days_since_ceph, days_since_mac(mac yes, no), days_since_phen(0-60 days, >=60 days, no treatment), days_since_tet(0-60 days, >=60 days, no treatment), -days_since_sul, days_since_flr(flr yes/no)## Farm season brd dan enr flr gam pen spc tet
## A:11 summer:43 neg:46 S:97 S:97 S:97 S:93 S:94 S:29 S:36
## B:11 winter:54 pos:51 R: 0 R: 0 R: 0 R: 4 R: 3 R:68 R:61
## C:20
## D:15
## E:32
## F: 8
## tild tilm tul total_abx last_treat days_since_pen
## S:82 S:86 S:93 0 : 8 <16 days : 9 <16 days : 4
## R:15 R:11 R: 4 1 to 2 :27 16-60 days:17 16-60 days: 6
## 3 or more:30 >=60 days :30 >=60 days : 8
## NA's :32 no trx : 9 no trx :47
## NA's :32 NA's :32
##
## days_since_ceph days_since_mac days_since_phen days_since_tet
## <16 days : 0 <16 days : 0 <16 days : 4 16-60 days: 1
## 16-60 days: 2 16-60 days: 2 16-60 days: 8 >=60 days :12
## >=60 days : 3 >=60 days :16 >=60 days :26 no trx :52
## no trx :60 no trx :47 no trx :27 NA's :32
## NA's :32 NA's :32 NA's :32
##
## days_since_sul days_since_flr
## >=60 days: 7 treated : 6
## no trx :58 >=60 days:11
## NA's :32 no trx :48
## NA's :32
##
## # remove variables
hsomind2 <- as.data.frame(hsom.ind) %>%
select(-"dan", -"enr", -"flr", -"gam", -"pen", -"tul",
-"days_since_ceph", -"days_since_sul")
# check
summary(hsomind2)## Farm season brd spc tet tild tilm total_abx
## A:11 summer:43 neg:46 S:29 S:36 S:82 S:86 0 : 8
## B:11 winter:54 pos:51 R:68 R:61 R:15 R:11 1 to 2 :27
## C:20 3 or more:30
## D:15 NA's :32
## E:32
## F: 8
## last_treat days_since_pen days_since_mac days_since_phen
## <16 days : 9 <16 days : 4 <16 days : 0 <16 days : 4
## 16-60 days:17 16-60 days: 6 16-60 days: 2 16-60 days: 8
## >=60 days :30 >=60 days : 8 >=60 days :16 >=60 days :26
## no trx : 9 no trx :47 no trx :47 no trx :27
## NA's :32 NA's :32 NA's :32 NA's :32
##
## days_since_tet days_since_flr
## 16-60 days: 1 treated : 6
## >=60 days :12 >=60 days:11
## no trx :52 no trx :48
## NA's :32 NA's :32
##
## # recategorize feed_lane_cleaning
hsomind2$total_abx <- factor(hsom.raw$total_abx, order = FALSE,
levels = c(0, 1, 2),
labels = c(">=2", ">=2",
"3 or more"))
hsomind2$last_treat <- factor(hsom.raw$last_treat, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
hsomind2$days_since_pen <- factor(hsom.raw$days_since_pen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
hsomind2$days_since_mac <- factor(hsom.raw$days_since_mac, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
hsomind2$days_since_phen <- factor(hsom.raw$days_since_phen, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("<60 days", "<60 days",
">=60 days", "no treatment"))
hsomind2$days_since_tet <- factor(hsom.raw$days_since_tet, order = FALSE,
levels = c(2, 3, 4),
labels = c("treated", "treated",
"no treatment"))
hsomind2$days_since_flr <- factor(hsom.raw$days_since_flr, order = FALSE,
levels = c(1, 2, 3, 4),
labels = c("treated", "treated",
"treated", "no treatment"))
summary(hsomind2)## Farm season brd spc tet tild tilm total_abx
## A:11 summer:43 neg:46 S:29 S:36 S:82 S:86 >=2 :35
## B:11 winter:54 pos:51 R:68 R:61 R:15 R:11 3 or more:30
## C:20 NA's :32
## D:15
## E:32
## F: 8
## last_treat days_since_pen days_since_mac days_since_phen
## <60 days :26 treated :18 treated :18 <60 days :12
## >=60 days :30 no treatment:47 no treatment:47 >=60 days :26
## no treatment: 9 NA's :32 NA's :32 no treatment:27
## NA's :32 NA's :32
##
##
## days_since_tet days_since_flr
## treated :13 treated :17
## no treatment:52 no treatment:48
## NA's :32 NA's :32
##
##
## # remove Farm E, tilmicosin, last_treatment
hsomind2$Farm <- factor(hsom.raw$farm, order = FALSE,
levels = c(1, 2, 3, 4, 6),
labels = c("A", "B", "C", "D", "F"))
summary(hsomind2)## Farm season brd spc tet tild tilm total_abx
## A :11 summer:43 neg:46 S:29 S:36 S:82 S:86 >=2 :35
## B :11 winter:54 pos:51 R:68 R:61 R:15 R:11 3 or more:30
## C :20 NA's :32
## D :15
## F : 8
## NA's:32
## last_treat days_since_pen days_since_mac days_since_phen
## <60 days :26 treated :18 treated :18 <60 days :12
## >=60 days :30 no treatment:47 no treatment:47 >=60 days :26
## no treatment: 9 NA's :32 NA's :32 no treatment:27
## NA's :32 NA's :32
##
##
## days_since_tet days_since_flr
## treated :13 treated :17
## no treatment:52 no treatment:48
## NA's :32 NA's :32
##
##
## hsomind.analyze <- na.omit(hsomind2) %>%
select(-"tilm", -"last_treat")
# check
summary(hsomind.analyze)## Farm season brd spc tet tild total_abx
## A:11 summer:27 neg:29 S:22 S:23 S:50 >=2 :35
## B:11 winter:38 pos:36 R:43 R:42 R:15 3 or more:30
## C:20
## D:15
## F: 8
## days_since_pen days_since_mac days_since_phen
## treated :18 treated :18 <60 days :12
## no treatment:47 no treatment:47 >=60 days :26
## no treatment:27
##
##
## days_since_tet days_since_flr
## treated :13 treated :17
## no treatment:52 no treatment:48
##
##
## hsomind <- hsomind.analyze %>%
summarise_each(funs())# create levels
hsom.bl.ind = tiers2blacklist(list("season", "Farm",
c("brd", "spc", "tet", "tild",
"total_abx", "days_since_pen",
"days_since_mac", "days_since_phen",
"days_since_tet", "days_since_flr")))
hsom.bl.ind = rbind(hsom.bl.ind, c("season"))
hsom.bl.ind## from to
## [1,] "Farm" "season"
## [2,] "brd" "season"
## [3,] "spc" "season"
## [4,] "tet" "season"
## [5,] "tild" "season"
## [6,] "total_abx" "season"
## [7,] "days_since_pen" "season"
## [8,] "days_since_mac" "season"
## [9,] "days_since_phen" "season"
## [10,] "days_since_tet" "season"
## [11,] "days_since_flr" "season"
## [12,] "brd" "Farm"
## [13,] "spc" "Farm"
## [14,] "tet" "Farm"
## [15,] "tild" "Farm"
## [16,] "total_abx" "Farm"
## [17,] "days_since_pen" "Farm"
## [18,] "days_since_mac" "Farm"
## [19,] "days_since_phen" "Farm"
## [20,] "days_since_tet" "Farm"
## [21,] "days_since_flr" "Farm"
## [22,] "season" "season"3.4 BNA
# individual factors hc structure learning
hsomind.hc <- hc(hsomind.analyze, blacklist = hsom.bl.ind, whitelist = wl, score = "aic")
hsomind.hc##
## Bayesian network learned via Score-based methods
##
## model:
## [season][Farm|season][brd|Farm][spc|Farm][total_abx|Farm]
## [days_since_tet|Farm:season][tet|Farm:brd][tild|season:brd]
## [days_since_pen|Farm:total_abx][days_since_mac|total_abx]
## [days_since_phen|total_abx:days_since_pen][days_since_flr|tet:total_abx]
## nodes: 12
## arcs: 17
## undirected arcs: 0
## directed arcs: 17
## average markov blanket size: 3.17
## average neighbourhood size: 2.83
## average branching factor: 1.42
##
## learning algorithm: Hill-Climbing
## score: AIC (disc.)
## penalization coefficient: 1
## tests used in the learning procedure: 264
## optimized: TRUE# plot customization
hsom.d <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomind.hc))
graph::nodeRenderInfo(hsom.d) <- list(fontsize=40)
Rgraphviz::renderGraph(hsom.d)
# individual factors tabu structure learning
hsomind.tabu <- tabu(hsomind.analyze, blacklist = hsom.bl.ind, whitelist = wl, score = "aic")
hsomind.tabu##
## Bayesian network learned via Score-based methods
##
## model:
## [season][Farm|season][brd|Farm][spc|Farm][total_abx|Farm]
## [days_since_tet|Farm:season][tet|Farm:brd][tild|season:brd]
## [days_since_pen|Farm:total_abx][days_since_mac|total_abx]
## [days_since_phen|total_abx:days_since_pen][days_since_flr|tet:total_abx]
## nodes: 12
## arcs: 17
## undirected arcs: 0
## directed arcs: 17
## average markov blanket size: 3.17
## average neighbourhood size: 2.83
## average branching factor: 1.42
##
## learning algorithm: Tabu Search
## score: AIC (disc.)
## penalization coefficient: 1
## tests used in the learning procedure: 484
## optimized: TRUE# plot customization
hsom.e <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomind.tabu))
graph::nodeRenderInfo(hsom.e) <- list(fontsize=40)
Rgraphviz::renderGraph(hsom.e)
graphviz.compare(hsomind.hc, hsomind.tabu, shape = "ellipse",
main= c("Hsom Individual Risk Factors HC DAG",
"Hsom Individual Risk Factors Tabu DAG"))
# bootstrap
hsomind.boot = boot.strength(hsomind.analyze, R = 10000, algorithm = "hc",
algorithm.args = list(blacklist = hsom.bl.ind,
whitelist = wl, score = "aic"))
# take arcs with strength of 50% and return average consensus network
hsomind.avg = averaged.network(hsomind.boot, threshold = 0.5)
# graph customization
hsom.f <- Rgraphviz::layoutGraph(bnlearn::as.graphNEL(hsomind.avg))
graph::nodeRenderInfo(hsom.f) <- list(fontsize=40)
Rgraphviz::renderGraph(hsom.f)
# plot the bootstrapped averaged plot with the original for comparison
graphviz.compare(hsomind.avg, hsomind.hc, shape = "ellipse",
main= c("Hsom Individual Factors Averaged DAG",
"Hsom Individual Factors Single DAG"))
3.5 Parameter Learning
Farm level query
# fit the parameters - bayesian method
hsomfarm.bayes = bn.fit(hsomfarm.avg, hsomall.analyze, method = "bayes")
# probability table
hsomfarm.bayes##
## Bayesian network parameters
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.4438776 0.5561224
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## source_no
## spc single low multi high multi
## S 0.007142857 0.723880597 0.251592357
## R 0.992857143 0.276119403 0.748407643
##
## Parameters of node tet (multinomial distribution)
##
## Conditional probability table:
## S R
## 0.372449 0.627551
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , bovishield = no
##
## season
## tild summer winter
## S 0.938356164 0.440594059
## R 0.061643836 0.559405941
##
## , , bovishield = yes
##
## season
## tild summer winter
## S 0.995049505 0.995726496
## R 0.004950495 0.004273504
##
##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.94848485 0.53225806
## R 0.05151515 0.46774194
##
## Parameters of node source_no (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## source_no A B C D E
## single 0.004975124 0.004975124 0.002754821 0.992673993 0.001727116
## low multi 0.990049751 0.990049751 0.002754821 0.003663004 0.001727116
## high multi 0.004975124 0.004975124 0.994490358 0.003663004 0.996545769
## Farm
## source_no F
## single 0.986394558
## low multi 0.006802721
## high multi 0.006802721
##
## Parameters of node onsite_milking (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## onsite_milking scrape lagoon clean sl
## no 0.996000000 0.006493506 0.008196721 0.003875969
## yes 0.004000000 0.993506494 0.991803279 0.996124031
##
## Parameters of node resp_vacc (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## resp_vacc scrape lagoon clean sl
## no 0.004000000 0.006493506 0.008196721 0.996124031
## yes 0.996000000 0.993506494 0.991803279 0.003875969
##
## Parameters of node salmonella (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## salmonella A B C D E
## no 0.007462687 0.007462687 0.004132231 0.928571429 0.997409326
## yes 0.992537313 0.992537313 0.995867769 0.071428571 0.002590674
## Farm
## salmonella F
## no 0.989795918
## yes 0.010204082
##
## Parameters of node bovishield (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bovishield A B C D E
## no 0.007462687 0.007462687 0.995867769 0.994505495 0.002590674
## yes 0.992537313 0.992537313 0.004132231 0.005494505 0.997409326
## Farm
## bovishield F
## no 0.989795918
## yes 0.010204082
##
## Parameters of node pyramid (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pyramid A B C D E F
## no 0.992537313 0.992537313 0.004132231 0.994505495 0.997409326 0.010204082
## yes 0.007462687 0.007462687 0.995867769 0.005494505 0.002590674 0.989795918
##
## Parameters of node titanium (multinomial distribution)
##
## Conditional probability table:
##
## feed_lane_cleaning
## titanium scrape lagoon clean sl
## no 0.996000000 0.993506494 0.008196721 0.996124031
## yes 0.004000000 0.006493506 0.991803279 0.003875969
##
## Parameters of node pinkeye (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## pinkeye A B C D E F
## no 0.992537313 0.007462687 0.004132231 0.005494505 0.997409326 0.010204082
## yes 0.007462687 0.992537313 0.995867769 0.994505495 0.002590674 0.989795918
##
## Parameters of node clostrid (multinomial distribution)
##
## Conditional probability table:
##
## titanium
## clostrid no yes
## no 0.996969697 0.016129032
## yes 0.003030303 0.983870968
##
## Parameters of node colost_source (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## colost_source no yes
## p 0.007692308 0.980916031
## dp 0.992307692 0.019083969
##
## Parameters of node colost_past (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## colost_past no yes
## no 0.992307692 0.003816794
## yes 0.007692308 0.996183206
##
## Parameters of node hosp_milk_pastuerized (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## hosp_milk_pastuerized no yes
## no 0.992307692 0.003816794
## yes 0.007692308 0.996183206
##
## Parameters of node action (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## action A B C D E
## screen 0.992537313 0.992537313 0.004132231 0.994505495 0.997409326
## premium 0.007462687 0.007462687 0.995867769 0.005494505 0.002590674
## Farm
## action F
## screen 0.989795918
## premium 0.010204082
##
## Parameters of node bull_calves (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## bull_calves A B C D E
## no 0.007462687 0.992537313 0.995867769 0.005494505 0.997409326
## yes 0.992537313 0.007462687 0.004132231 0.994505495 0.002590674
## Farm
## bull_calves F
## no 0.010204082
## yes 0.989795918
##
## Parameters of node milk_mass_med (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## milk_mass_med A B C D E
## no 0.007462687 0.992537313 0.995867769 0.994505495 0.002590674
## yes 0.992537313 0.007462687 0.004132231 0.005494505 0.997409326
## Farm
## milk_mass_med F
## no 0.010204082
## yes 0.989795918
##
## Parameters of node mass_pen_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_pen_exp A B C D E
## no 0.007462687 0.992537313 0.995867769 0.994505495 0.997409326
## yes 0.992537313 0.007462687 0.004132231 0.005494505 0.002590674
## Farm
## mass_pen_exp F
## no 0.989795918
## yes 0.010204082
##
## Parameters of node mass_rum_exp (multinomial distribution)
##
## Conditional probability table:
##
## mass_pen_exp
## mass_rum_exp no yes
## no 0.002890173 0.978260870
## yes 0.997109827 0.021739130
##
## Parameters of node mass_corr_exp (multinomial distribution)
##
## Conditional probability table:
##
## bull_calves
## mass_corr_exp no yes
## no 0.003937008 0.992753623
## yes 0.996062992 0.007246377
##
## Parameters of node mass_sul_exp (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## mass_sul_exp A B C D E
## no 0.007462687 0.007462687 0.004132231 0.005494505 0.997409326
## yes 0.992537313 0.992537313 0.995867769 0.994505495 0.002590674
## Farm
## mass_sul_exp F
## no 0.989795918
## yes 0.010204082
##
## Parameters of node feed_lane_cleaning (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## feed_lane_cleaning A B C D E
## scrape 0.988805970 0.003731343 0.993801653 0.002747253 0.001295337
## lagoon 0.003731343 0.988805970 0.002066116 0.002747253 0.001295337
## clean 0.003731343 0.003731343 0.002066116 0.991758242 0.001295337
## sl 0.003731343 0.003731343 0.002066116 0.002747253 0.996113990
## Farm
## feed_lane_cleaning F
## scrape 0.005102041
## lagoon 0.984693878
## clean 0.005102041
## sl 0.005102041
##
## Parameters of node dust_mgmt (multinomial distribution)
##
## Conditional probability table:
##
## resp_vacc
## dust_mgmt no yes
## no 0.992307692 0.003816794
## yes 0.007692308 0.996183206
##
## Parameters of node no_classes (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## no_classes A B C D E
## 0 0.363184080 0.184079602 0.002754821 0.069597070 0.996545769
## 1 0.631840796 0.363184080 0.647382920 0.465201465 0.001727116
## 2 or more 0.004975124 0.452736318 0.349862259 0.465201465 0.001727116
## Farm
## no_classes F
## 0 0.251700680
## 1 0.496598639
## 2 or more 0.251700680
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.208812261 0.038226300
## B 0.001915709 0.203363914
## C 0.231800766 0.185015291
## D 0.070881226 0.221712538
## E 0.369731801 0.295107034
## F 0.116858238 0.056574924# conditional probabilities
hsomfarm.bayes$tilm##
## Parameters of node tilm (multinomial distribution)
##
## Conditional probability table:
##
## tild
## tilm S R
## S 0.94848485 0.53225806
## R 0.05151515 0.46774194hsomfarm.bayes$tet##
## Parameters of node tet (multinomial distribution)
##
## Conditional probability table:
## S R
## 0.372449 0.627551hsomfarm.bayes$spc##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## source_no
## spc single low multi high multi
## S 0.007142857 0.723880597 0.251592357
## R 0.992857143 0.276119403 0.748407643hsomfarm.bayes$gen## NULLset.seed(118)
cpquery(hsomfarm.bayes, (tild =="R"), bovishield == "yes")## [1] 0.004460303cpquery(hsomfarm.bayes, (tild =="R"), bovishield == "no")## [1] 0.3599736cpquery(hsomfarm.bayes, (tild =="S"), bovishield == "yes")## [1] 0.994639cpquery(hsomfarm.bayes, (tild =="S"), bovishield == "no")## [1] 0.6594837set.seed(119)
cpquery(hsomfarm.bayes, (tilm == "R"), tild == "R")## [1] 0.4666247cpquery(hsomfarm.bayes, (tilm == "S"), tild == "R")## [1] 0.5251656cpquery(hsomfarm.bayes, (tild == "R"), tilm == "R")## [1] 0.627551cpquery(hsomfarm.bayes, (tild == "S"), tilm == "S")## [1] 0.9038243set.seed(120)
cpquery(hsomfarm.bayes, (tilm == "R"), tild == "R" & bovishield == "yes")## [1] 0.4705882cpquery(hsomfarm.bayes, (tilm == "R"), tild == "S" & bovishield == "yes")## [1] 0.04875887cpquery(hsomfarm.bayes, (tilm == "R"), tild == "S" & bovishield == "no")## [1] 0.04925939cpquery(hsomfarm.bayes, (tilm == "S"), tild == "R" & bovishield == "yes")## [1] 0.6538462cpquery(hsomfarm.bayes, (tilm == "S"), tild == "S" & bovishield == "no")## [1] 0.9498056Animal level query
# fit the parameters - bayesian method
hsomind.bayes = bn.fit(hsomind.hc, hsomind.analyze, method = "bayes")
# probability table
hsomind.bayes##
## Bayesian network parameters
##
## Parameters of node Farm (multinomial distribution)
##
## Conditional probability table:
##
## season
## Farm summer winter
## A 0.330909091 0.054545455
## B 0.003636364 0.288311688
## C 0.367272727 0.262337662
## D 0.112727273 0.314285714
## F 0.185454545 0.080519481
##
## Parameters of node season (multinomial distribution)
##
## Conditional probability table:
## summer winter
## 0.4166667 0.5833333
##
## Parameters of node brd (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## brd A B C D F
## neg 0.455357143 0.008928571 0.648514851 0.401315789 0.621951220
## pos 0.544642857 0.991071429 0.351485149 0.598684211 0.378048780
##
## Parameters of node spc (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## spc A B C D F
## S 0.633928571 0.812500000 0.301980198 0.006578947 0.012195122
## R 0.366071429 0.187500000 0.698019802 0.993421053 0.987804878
##
## Parameters of node tet (multinomial distribution)
##
## Conditional probability table:
##
## , , brd = neg
##
## Farm
## tet A B C D F
## S 0.205882353 0.500000000 0.614503817 0.008196721 0.794117647
## R 0.794117647 0.500000000 0.385496183 0.991803279 0.205882353
##
## , , brd = pos
##
## Farm
## tet A B C D F
## S 0.008196721 0.184684685 0.147887324 0.554945055 0.661290323
## R 0.991803279 0.815315315 0.852112676 0.445054945 0.338709677
##
##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , brd = neg
##
## season
## tild summer winter
## S 0.91509434 0.37692308
## R 0.08490566 0.62307692
##
## , , brd = pos
##
## season
## tild summer winter
## S 0.99122807 0.81460674
## R 0.00877193 0.18539326
##
##
## Parameters of node total_abx (multinomial distribution)
##
## Conditional probability table:
##
## Farm
## total_abx A B C D F
## >=2 0.991071429 0.366071429 0.450495050 0.532894737 0.378048780
## 3 or more 0.008928571 0.633928571 0.549504950 0.467105263 0.621951220
##
## Parameters of node days_since_pen (multinomial distribution)
##
## Conditional probability table:
##
## , , total_abx = >=2
##
## Farm
## days_since_pen A B C D F
## treated 0.004504505 0.012195122 0.445054945 0.006172840 0.016129032
## no treatment 0.995495495 0.987804878 0.554945055 0.993827160 0.983870968
##
## , , total_abx = 3 or more
##
## Farm
## days_since_pen A B C D F
## treated 0.500000000 0.852112676 0.725225225 0.007042254 0.009803922
## no treatment 0.500000000 0.147887324 0.274774775 0.992957746 0.990196078
##
##
## Parameters of node days_since_mac (multinomial distribution)
##
## Conditional probability table:
##
## total_abx
## days_since_mac >=2 3 or more
## treated 0.0915493 0.5000000
## no treatment 0.9084507 0.5000000
##
## Parameters of node days_since_phen (multinomial distribution)
##
## Conditional probability table:
##
## , , days_since_pen = treated
##
## total_abx
## days_since_phen >=2 3 or more
## <60 days 0.01960784 0.35672515
## >=60 days 0.01960784 0.21637427
## no treatment 0.96078431 0.42690058
##
## , , days_since_pen = no treatment
##
## total_abx
## days_since_phen >=2 3 or more
## <60 days 0.13066667 0.18974359
## >=60 days 0.35466667 0.74358974
## no treatment 0.51466667 0.06666667
##
##
## Parameters of node days_since_tet (multinomial distribution)
##
## Conditional probability table:
##
## , , season = summer
##
## Farm
## days_since_tet A B C D F
## treated 0.005494505 0.500000000 0.004950495 0.983870968 0.009803922
## no treatment 0.994505495 0.500000000 0.995049505 0.016129032 0.990196078
##
## , , season = winter
##
## Farm
## days_since_tet A B C D F
## treated 0.023809524 0.004504505 0.004950495 0.582644628 0.983870968
## no treatment 0.976190476 0.995495495 0.995049505 0.417355372 0.016129032
##
##
## Parameters of node days_since_flr (multinomial distribution)
##
## Conditional probability table:
##
## , , total_abx = >=2
##
## tet
## days_since_flr S R
## treated 0.109756098 0.004950495
## no treatment 0.890243902 0.995049505
##
## , , total_abx = 3 or more
##
## tet
## days_since_flr S R
## treated 0.311320755 0.702898551
## no treatment 0.688679245 0.297101449hsomind.bayes$tild##
## Parameters of node tild (multinomial distribution)
##
## Conditional probability table:
##
## , , brd = neg
##
## season
## tild summer winter
## S 0.91509434 0.37692308
## R 0.08490566 0.62307692
##
## , , brd = pos
##
## season
## tild summer winter
## S 0.99122807 0.81460674
## R 0.00877193 0.18539326set.seed(121)
cpquery(hsomind.bayes, (tet == "R"), brd == "neg")## [1] 0.5433954cpquery(hsomind.bayes, (tet == "R"), brd == "pos")## [1] 0.715339cpquery(hsomind.bayes, (tet == "S"), brd == "neg")## [1] 0.4572265cpquery(hsomind.bayes, (tet == "S"), brd == "pos")## [1] 0.2833515set.seed(122)
cpquery(hsomind.bayes, (tet == "R"), brd == "neg" & days_since_phen == "no treatment")## [1] 0.5576037cpquery(hsomind.bayes, (tet == "R"), brd == "pos" & days_since_phen == "no treatment")## [1] 0.7771976cpquery(hsomind.bayes, (tet == "S"), brd == "neg" & days_since_phen == "no treatment")## [1] 0.4358407cpquery(hsomind.bayes, (tet == "S"), brd == "pos" & days_since_phen == "no treatment")## [1] 0.2343324set.seed(123)
cpquery(hsomind.bayes, (tet == "R"), brd == "neg" & days_since_phen == "<60 days")## [1] 0.4766977cpquery(hsomind.bayes, (tet == "R"), brd == "pos" & days_since_phen == "<60 days")## [1] 0.7255639cpquery(hsomind.bayes, (tet == "S"), brd == "neg" & days_since_phen == "<60 days")## [1] 0.5037594cpquery(hsomind.bayes, (tet == "S"), brd == "pos" & days_since_phen == "<60 days")## [1] 0.2771996set.seed(124)
cpquery(hsomind.bayes, (tet == "R"), brd == "neg" & days_since_phen == ">=60 days")## [1] 0.5581662cpquery(hsomind.bayes, (tet == "R"), brd == "pos" & days_since_phen == ">=60 days")## [1] 0.652515cpquery(hsomind.bayes, (tet == "S"), brd == "neg" & days_since_phen == ">=60 days")## [1] 0.4328439cpquery(hsomind.bayes, (tet == "S"), brd == "pos" & days_since_phen == ">=60 days")## [1] 0.35691474. Summary Statistics
# AMR
amr <- rbind.fill(pmult.all, mhaem.all, hsom.all)
summary(amr)## Farm season dan enr flr gam gen
## A:47 summer:174 S : 91 S : 89 S :138 S:209 S :251
## B:61 winter:187 R :173 R :175 R :126 R:152 R : 13
## C:58 NA's: 97 NA's: 97 NA's: 97 NA's: 97
## D:69
## E:65
## F:61
## pen spc tild tilm tul source_no onsite_milking
## S:310 S:239 S:173 S:162 S:297 single :130 no :105
## R: 51 R:122 R:188 R:199 R: 64 low multi :108 yes:256
## high multi:123
##
##
##
## resp_vacc first_def salmonella bovishield pyramid titanium pinkeye
## no : 65 no :211 no :194 no :188 no :242 no :292 no :112
## yes:296 yes : 53 yes:167 yes:173 yes:119 yes: 69 yes:249
## NA's: 97
##
##
##
## clostrid colost_source colost_past colost_again hosp_milk_pastuerized
## no :292 p :295 no : 65 no : 69 no : 65
## yes: 69 dp: 66 yes:296 yes : 88 yes:296
## -99 :107
## NA's: 97
##
##
## action bull_calves milk_mass_med mass_pen_exp mass_tet_exp mass_rum_exp
## screen :303 no :184 no :188 no :314 no : 61 no : 47
## premium: 58 yes:177 yes:173 yes: 47 yes:300 yes:314
##
##
##
##
## mass_corr_exp mass_sul_exp feed_lane_cleaning dust_mgmt no_classes
## no :177 no :126 scrape:105 no : 65 0 :124
## yes:184 yes:235 lagoon:122 yes:296 1 :146
## clean : 69 2 or more: 65
## sl : 65 2 : 26
## 3 or more: 0
##
## cef tet
## S :119 S : 52
## R : 0 R :164
## NA's:242 NA's:145
##
##
## # animal-level
all_calves <- rbind.fill(pmultind.analyze, mhaemind.analyze, hsomind.analyze) %>%
select(Farm, brd, total_abx, days_since_pen, days_since_mac, days_since_phen,
days_since_flr, days_since_tet, days_since_ceph, tild, tet, spc, pen,
flr, tilm, enr, dan, gam, tul) %>%
distinct()
summary(all_calves)## Farm brd total_abx days_since_pen days_since_mac
## A:33 neg:126 0 : 35 <60 days : 16 <60 days : 12
## B:58 pos:137 1 to 2 : 63 >=60 days : 10 >=60 days : 26
## C:57 3 or more:134 no treatment:203 no treatment:185
## D:60 >=2 : 31 treated : 34 treated : 40
## F:55
## days_since_phen days_since_flr days_since_tet
## <60 days : 36 <60 days : 13 treatment : 21
## >=60 days : 94 >=60 days : 42 no treatment:217
## no treatment:110 no treatment:178 treated : 25
## 0-60 days : 23 0-60 days : 13
## treated : 17
## days_since_ceph tild tet spc pen flr
## <60 days : 10 S:117 S : 35 S:177 S : 59 S :103
## >=60 days : 10 R:146 R :120 R: 86 R : 36 R :100
## no treatment:174 NA's:108 NA's:168 NA's: 60
## treated : 9
## NA's : 60
## tilm enr dan gam tul
## S : 56 S : 73 S : 74 S : 87 S :150
## R :147 R :130 R :129 R :116 R : 53
## NA's: 60 NA's: 60 NA's: 60 NA's: 60 NA's: 60
##
## summary(pmultind.analyze)## Farm season brd dan enr flr gam spc tild tilm
## A:13 summer:61 neg:63 S:51 S:51 S:49 S:37 S:88 S:32 S:29
## B:24 winter:60 pos:58 R:70 R:70 R:72 R:84 R:33 R:89 R:92
## C:21
## D:34
## F:29
## tul total_abx last_treat days_since_pen days_since_ceph
## S:98 0 :21 <16 days :19 <60 days :17 <60 days : 10
## R:23 1 to 2 :34 16-60 days:33 >=60 days :10 >=60 days : 10
## 3 or more:66 >=60 days :48 no treatment:94 no treatment:101
## no trx :21
##
## days_since_mac days_since_phen days_since_tet days_since_sul
## <60 days :12 <60 days :27 treatment :25 >=60 days: 20
## >=60 days :27 >=60 days :42 no treatment:96 no trx :101
## no treatment:82 no treatment:52
##
##
## days_since_flr
## <60 days :14
## >=60 days :29
## no treatment:78
##
## summary(mhaemind.analyze)## Farm season brd dan enr flr gam pen spc tet
## A:23 summer:55 neg:50 S:35 S:34 S:72 S:71 S:70 S:93 S:15
## B:26 winter:55 pos:60 R:75 R:76 R:38 R:39 R:40 R:17 R:95
## C:17
## D:20
## F:24
## tild tilm tul total_abx last_treat days_since_pen
## S:55 S:42 S:78 0 :30 <16 days :14 treated :18
## R:55 R:68 R:32 1 to 2 :35 16-60 days:25 no treatment:92
## 3 or more:45 >=60 days :42
## no trx :29
##
## days_since_ceph days_since_mac days_since_phen
## treated : 10 treated :24 0-60 days :24
## no treatment:100 no treatment:86 >=60 days :34
## no treatment:52
##
##
## days_since_tet days_since_flr
## treated :15 0-60 days :13
## no treatment:95 >=60 days :16
## no treatment:81
##
## summary(hsomind.analyze)## Farm season brd spc tet tild total_abx
## A:11 summer:27 neg:29 S:22 S:23 S:50 >=2 :35
## B:11 winter:38 pos:36 R:43 R:42 R:15 3 or more:30
## C:20
## D:15
## F: 8
## days_since_pen days_since_mac days_since_phen
## treated :18 treated :18 <60 days :12
## no treatment:47 no treatment:47 >=60 days :26
## no treatment:27
##
##
## days_since_tet days_since_flr
## treated :13 treated :17
## no treatment:52 no treatment:48
##
##
##