Search Results (26)

This study examines the effectiveness of Generalised Additive Models (GAMs) and log-log linear models for estimating the parameters of the generalised extreme value (GEV) distribution, which are then used to estimate flood quantiles in ungauged catchments. This is known as the parameter regression technique (PRT). Using data from 88 gauged catchments in New South Wales, Australia, flood quantiles were estimated for various annual exceedance probabilities (AEPs) of 50%, 20%, 10%, 5%, 2%, and 1%, corresponding to return periods of 2, 5, 10, 20, 50, and 100 years, denoted by Q₂, Q₅, Q₁₀, Q₂₀, Q₅₀, and Q₁₀₀, respectively. These flood quantiles were then used as dependent variables, while several catchment characteristics served as independent variables in the regression. GAMs were employed to capture non-linearities in flood generation processes. This study evaluates different GAMs and log-log linear models, identifying the best ones based on significant predictors and various statistical metrics using a leave-one-out (LOO) validation approach. The results indicate that GAMs provide more accurate and reliable predictions of flood quantiles compared to the log-log linear models, demonstrating better performance in capturing observed values across different quantiles. The absolute median relative error percentage (REr%) ranges from 33% to 39% for the GAMs and from 36% to 45% for the log-log models. GAMs demonstrate better performance compared to the log-log linear models for quantiles Q₂, Q₅, Q₁₀, Q₂₀, and Q₅₀; however, their performances appear to be similar for Q₁₀₀. Full article

Multispectral modelling of 114 tropical Andean lakes in Southern Ecuador was implemented using observations of the maximum depth (Z_max). Five machine learning methods (MLMs), namely the multiple linear regression model (MLRM), generalised additive model (GAM), generalised linear model (GLM), multivariate adaptive regression splines (MARS), and random forest (RF), were applied on a LANDSAT 8 mosaic. Within the scope of a split-sample (SS) evaluation test, for each of the MLMs, a single model was developed for 70% (i.e., 80) of the studied lakes. Statistical measures and graphical inspection were used in the evaluation tests. An analysis of the absolute value of the model residuals (|res|) revealed that the MARS method outperformed the other MLMs. Nevertheless, a |res| > 10 m was observed for approximately 10% of the lakes. The worst predictions were produced by the GLM. These findings were confirmed in the model validation phase (SS test). With the exception of the GLM, the MLMs correctly predicted whether a lake was shallow or deep in more than 80% of the cases. In a more stringent multi-site (MS) test, the performance of the five Z_max models was assessed in predicting the bathymetry of 11,636 pixels that were not considered when fitting the models. Once more, MARS outperformed the other MLMs. However, a |res| > 10 m for 20% of the pixels was observed. Nevertheless, the quality of the predictions may still be regarded as acceptable for management purposes. Promising multispectral bathymetric predictions could be obtained, even with only a limited number of observations. The evaluation tests used in this pioneering study could be easily replicated elsewhere. Full article

Background: Despite the implementation of various preventive measures, India continues to experience an alarmingly high under-five mortality rate (U5MR). The most recent nationwide data on U5MRs has provided an opportunity to re-examine the associated factors of U5MRs using advanced techniques. This study attempted to identify the associated determinants of U5MRs via the generalised additive Cox proportional hazards method. Methods: This study analysed the fifth round of unit-level data for 213,612 children from the National Family Health Survey (NFHS-5) to identify the risk factors associated with U5MRs, employing a generalised additive Cox proportional hazards regression analysis. Results: The children who had a length of pregnancy of less than 9 months had a 2.621 (95% CI: 2.494, 2.755) times greater hazard of U5MRs than the children who had a gestational period of 9 months or more. The non-linear association with U5MRs was highest in the mother’s age, followed by the mother’s haemoglobin, the mother’s education, and household wealth score. The relationships between the mother’s age and the mother’s haemoglobin level with the U5MR were found to be U-shaped. Conclusions: This study highlights the importance of addressing maternal and socioeconomic factors while improving access to healthcare services in order to reduce U5MRs in India. Furthermore, the findings underscore the necessity for more sophisticated approaches to healthcare delivery that consider the non-linear relationships between predictor variables and U5MRs. Full article

Changes in water quality are closely linked to seasonal fluctuations in streamflow, and a thorough understanding of how these variations interact across different time scales is important for the efficient management of surface water bodies such as rivers, lakes, and reservoirs. The aim of this study is to explore the potential connection between streamflow, rainfall, and water quality and propose an optimised ensemble model for the prediction of a water quality index (WQI). This study modelled the changes in five water quality parameters such as ammonia nitrogen (NH₃-N), phosphate (PO₄³⁻), pH, turbidity, total dissolved solids (TDS), and their associated WQI caused by rainfall and streamflow. The analysis was conducted across three temporal scales, weekly, monthly, and seasonal, using a generalised additive model (GAM) in Toowoomba, Australia. TDS, turbidity, and WQI exhibited a significant nonlinear variation with the changes in streamflow in the weekly and monthly scales. Additionally, pH demonstrated a significant linear to weakly linear correlation with discharge across the three temporal scales. For the accurate prediction of WQI, this study proposed an ensemble model integrating an extreme gradient boosting (XGBoost) and Bayesian optimisation (BO) algorithm, using streamflow as an input across the same temporal scales. The results for the three temporal scales provided the best accuracy of monthly data, based on the accuracy metrics R² (0.91), MAE (0.20), and RMSE (0.42). The comparison between the test and predicted data indicated that the prediction model overestimated the WQI at some points. This study highlights the efficiency of integrating rainfall, streamflow, and water quality correlations for WQI prediction, which can provide valuable insights for guiding future water management strategies in similar catchment areas, especially amidst changing climatic conditions. Full article

Air pollutants are known to trigger asthma and wheezing-associated lower respiratory infections in children, but evidence regarding their effect on hospital admissions in India is limited. We conducted a time-series study over a period of five years to assess the role of ambient air pollutants in daily asthma-related hospital admissions in children in Mysore, India. Daily asthma and wheeze (associated with lower respiratory infections) admissions were modelled using a generalised additive model (GAM) to examine the non-linear effects and generalised linear models (GLM) for linear effects, if any. Models were adjusted by day of the week and lag days, with smooth terms for time, maximum temperature, and relative humidity, and they were stratified by sex and age group. Of the 362 children admitted, more than 50% were boys, and the mean age was 5.34 years (±4.66). The GAMs showed non-linear associations between NO₂, PM_2.5, and NH₃. For example, a 10 µgm⁻³ (or 10%) increase in NO₂ increased admissions by 2.42. These non-linear effects were more pronounced in boys. A linear effect was detected for PM₁₀ with a relative risk (95% CI) of 1.028, 1.013, and 1.043 with admission. Further research is needed to explore whether these findings can be replicated in different cities in India. Air pollution needs to be controlled, and policies that focus on lower cut-off levels for vulnerable populations are necessary. Full article

Some retrospective studies suggest that psychosocial stressors trigger the onset of tics. This study examined prospective hypothalamic–pituitary–adrenal (HPA) axis activity and perceived stress prior to tic onset. In the present study, 259 children at high risk for developing tics were assessed for hair cortisol concentration (HCC) and parent-on-child-reported perceived stress four-monthly over a three-year period. We used (i) generalised additive modelling (GAM) to investigate the time effects on HCC (hair samples n = 765) and perceived stress (questionnaires n = 1019) prior to tic onset and (ii) binary logistic regression to predict tic onset in a smaller subsample with at least three consecutive assessments (six to nine months before, two to five months before, and at tic onset). GAM results indicated a non-linear increasing course of HCC in children who developed tics, and a steady HCC course in those without tics, as well as a linear-increasing course of perceived stress in both groups. Logistic regression showed that with a higher HCC in hair samples collected in a range of two to five months before tic onset (which refers to cortisol exposure in a range of four to eight months), the relative likelihood of tic onset rose. Our study suggests increased stress prior to tic onset, as evidenced by higher HCC several months before tic onset. Full article

The outbreaks of two strains of rabbit haemorrhagic disease (RHD) (GI.1 and GI.2) in the Iberian Peninsula have caused substantial economic losses in commercial rabbitries and have affected the conservation of rabbit-sensitive predators due to the dramatic decline of their natural populations. However, the assessment of the impact of both RHD strains on wild rabbit populations has been limited to a few small-scale studies. Little is known about the overall impact within its native range. In this study, we described and compared the effects of GI.1 and GI.2 countrywide by using time series of hunting bag data widely available across the country and compared their trend during the first eight years after the first outbreak of GI.1 (i.e., 1998) and GI.2 (i.e., 2011), respectively. We used Gaussian generalised additive models (GAM) with the number of hunted rabbits as the response variable and year as the predictor to evaluate the non-linear temporal dynamics of the population at the national and regional community levels. The first GI.1 caused a population decline of around 53%, affecting most Spanish regional communities where the disease occurred. The positive trend observed after GI.1 in Spain ended with the initial outbreak of GI.2, which did not appear to cause a national population decline. In contrast, we found significant variability in the rabbit population trend among regional communities, where some increased, and others decreased. Such a disparity is unlikely to be explained by a single factor; rather, it appears to result from several factors, such as climatic conditions, host resistance improvement, virulence attenuation, or population density. Our study suggests that a national comprehensive hunting bag series could aid in elucidating the differences in the impact of emerging diseases on a large scale. Future research should focus on national longitudinal serological studies to shed light on the immunological status of rabbit populations in different regions to better understand the evolution of RHD strains and the resistance gained by the wild populations. Full article

A lack of transparency in machine learning models can limit their application. We show that analysis of variance (ANOVA) methods extract interpretable predictive models from them. This is possible because ANOVA decompositions represent multivariate functions as sums of functions of fewer variables. Retaining the terms in the ANOVA summation involving functions of only one or two variables provides an efficient method to open black box classifiers. The proposed method builds generalised additive models (GAMs) by application of L1 regularised logistic regression to the component terms retained from the ANOVA decomposition of the logit function. The resulting GAMs are derived using two alternative measures, Dirac and Lebesgue. Both measures produce functions that are smooth and consistent. The term partial responses in structured models (PRiSM) describes the family of models that are derived from black box classifiers by application of ANOVA decompositions. We demonstrate their interpretability and performance for the multilayer perceptron, support vector machines and gradient-boosting machines applied to synthetic data and several real-world data sets, namely Pima Diabetes, German Credit Card, and Statlog Shuttle from the UCI repository. The GAMs are shown to be compliant with the basic principles of a formal framework for interpretability. Full article

Pricing using a Generalised Linear Model is the gold standard in the auto insurance industry and rate regulation. Generalised Additive Model applications in insurance pricing are receiving increasing attention from academic researchers and actuarial pricing professionals. The actuarial practice has constantly shown evidence of significantly different premium rates among the different rating territories. In this work, we build predictive models for claim frequency and severity using the synthetic Usage Based Insurance (UBI) dataset variables. First, we conduct territorial clustering based on each location’s claim counts and amounts by grouping those locations into a smaller set, defined as a cluster for rating purposes. After clustering, we incorporate these clusters into our predictive model to determine the risk relativity for each factor level. Through predictive modelling, we have successfully identified key factors that may be helpful for the rate regulation of UBI. Our work aims to fill the gap between individual-level pricing and rate regulation using the UBI database and provides insights on consistency in using traditional rating variables for UBI pricing. Our main contribution is to outline how GAM can address a more complicated functionality of risk factors and the interactions among them. We also contribute to demonstrating the territory clustering problem in UBI to construct the rating territories for pricing and rate regulation. We find that relativity for high annual mileage driven is almost three times that associated with low annual mileage level, which implies its importance in premium calculation. Overall, we provide insights into how UBI can be regulated through traditional pricing factors, additional factors from UBI datasets and rating territories derived from basic rating units and the driver’s location. Full article

Data-driven condition-based maintenance (CBM) and predictive maintenance (PdM) strategies have emerged over recent years and aim at minimizing the aviation maintenance costs and environmental impact by the diagnosis and prognosis of aircraft systems. As the use of data and relevant algorithms is essential to AI-based gas turbine diagnostics, there are different technical, operational, and regulatory challenges that need to be tackled in order for the aeronautical industry to be able to exploit their full potential. In this work, the machine learning (ML) method of the generalised additive model (GAM) is used in order to predict the evolution of an aero engine’s exhaust gas temperature (EGT). Three different continuous synthetic data sets developed by NASA are employed, known as New Commercial Modular Aero-Propulsion System Simulation (N-CMAPSS), with increasing complexity in engine deterioration. The results show that the GAM can be predict the evolution of the EGT with high accuracy when using several input features that resemble the types of physical sensors installed in aero gas turbines currently in operation. As the GAM offers good interpretability, this case study is used to discuss the different data attributes a data set needs to have in order to build trust and move towards certifiable models in the future. Full article

For design flood estimation in ungauged catchments, Regional Flood Frequency Analysis (RFFA) is commonly used. Most of the RFFA methods are primarily based on linear modelling approaches, which do not account for the inherent nonlinearity of rainfall-runoff processes. Using data from 114 catchments in Victoria, Australia, this study employs the Generalised Additive Model (GAM) in RFFA and compares the results with linear method known as Quantile Regression Technique (QRT). The GAM model performance is found to be better for smaller return periods (i.e., 2, 5 and 10 years) with a median relative error ranging 16–41%. For higher return periods (i.e., 20, 50 and 100 years), log-log linear regression model (QRT) outperforms the GAM model with a median relative error ranging 31–59%. Full article

This study quantifies the air quality impact on population mortality from an actuarial perspective, considering implications to the industry through the application of findings. The study focuses on the increase in mortality from air quality changes due to extreme weather impacts. We conduct an empirical study using monthly Californian climate and mortality data from 1999 to 2019 to determine whether adding PM2.5 as a factor improves forecast excess mortality. Expected mortality is defined using the rolling five-year average of observed mortality for each county. We compared three statistical models, namely a Generalised Linear Model (GLM), a Generalised Additive Model (GAM), and an Extreme Gradient Boosting (XGB) regression model. We find including PM2.5 improves the performance of all three models and that the GAM performs the best in terms of predictive accuracy. Change points are also considered to determine whether significant events trigger changes in mortality over extended periods. Based on several identified change points, some wildfires trigger heightened excess mortality. Full article

Gadiculus argenteus, is a quite common and relatively abundant fish present in the Galicia and Cantabrian Sea continental shelf and is one of the main trophic resources in the area. Despite its importance in the trophic ecosystem dynamics of the Spanish northern continental shelf, there is a general lack of knowledge about the ecological preferences of the species. The aims of this study are both to determine the importance of spatial, temporal, and oceanic environmental factors on the distribution of G. argenteus in this area and to generate, for the first time for the species, abundance maps that could help in the development of, for example, trophic models or marine management plans. In order to reach these goals, data on the abundance of this species from an annual bottom trawl survey (DEMERSALES) for the period 1998–2019, along with temporally invariant (depth, slope, sediment type, and percentage of organic matter) and annual (near-bottom temperature and salinity and chlorophyll-a concentration) environmental layers were modelled using delta generalised additive models (GAMs). The results helped us to identify the most suitable habitats for the species and which environmental factors have a significant effect on its distribution. According to our findings, the species showed higher abundances in the upper slope and a preference for muddy bottoms, with chlorophyll-a positively influencing its biomass. It aggregates mainly in the Galician waters and in the most eastern longitudes of the study area. The results of the models proved that most of the environmental variables chosen are relevant factors in the distribution of the species. Full article

Previous studies found that exposure to ambient nitrogen dioxide (NO₂) was associated with an increased risk of incident stroke, but few studies have been conducted for relatively low NO₂ pollution areas. In this study, the short-term effects of NO₂ on the risk of incident stroke in a relatively low-pollution area, Enshi city of Hubei Province, China, were investigated through time-series analysis. Daily air-pollution data, meteorological data, and stroke incidence data of residents in Enshi city from 1 January 2015 to 31 December 2018 were collected. A time-series analysis using a generalised additive model (GAM) based on Poisson distribution was applied to explore the short-term effects of low-level NO₂ exposure on the risk of incident stroke and stroke subtypes, as well as possible age, sex, and seasonal differences behind the effects. In the GAM model, potential confounding factors, such as public holidays, day of the week, long-term trends, and meteorological factors (temperature and relative humidity), were controlled. A total of 9122 stroke incident cases were included during the study period. We found that NO₂ had statistically significant effects on the incidence of stroke and ischemic stroke, estimated by excess risk (ER) of 0.37% (95% CI: 0.04–0.70%) and 0.58% (95% CI: 0.18–0.98%), respectively. For the cumulative lag effects, the NO₂ still had a statistically significant effect on incident ischemic stroke, estimated by ER of 0.61% (95% CI: 0.01–1.21%). The two-pollutant model showed that the effects of NO₂ on incident total stroke were still statistically significant after adjusting for other air pollutants (PM_2.5, PM₁₀, SO₂, CO, and O₃). In addition, the effects of NO₂ exposure on incident stroke were statistically significant in elderly (ER = 0.75%; 95% CI: 0.11–1.40%), males (ER = 0.47%; 95% CI: 0.05–0.89%) and cold season (ER = 0.83%; 95% CI: 0.15–1.51%) subgroups. Our study showed that, as commonly observed in high-pollution areas, short-term exposure to low-level NO₂ was associated with an increased risk of incident stroke, including ischemic stroke. Males and elderly people were more vulnerable to the effects of NO₂, and the adverse effects might be promoted in the cold season. Full article

In this paper, the emission sources of PM₁₀ are characterised by analysing its trace elements (TE) and ions contents. PM₁₀ samples were collected for a year (2019–2020) at five sites and analysed. PM₁₀ speciated data were analysed using graphical visualization, correlation analysis, generalised additive model (GAM), and positive matrix factorization (PMF). Annual average PM₁₀ concentrations (µg/m³) were 304.68 ± 155.56 at Aziziyah, 219.59 ± 87.29 at Misfalah, 173.90 ± 103.08 at Abdeyah, 168.81 ± 82.50 at Askan, and 157.60 ± 80.10 at Sanaiyah in Makkah, which exceeded WHO (15 µg/m³), USEPA (50 µg/m³), and the Saudi Arabia national (80 µg/m³) annual air quality standards. A GAM model was developed using PM₁₀ as a response and ions and TEs as predictors. Among the predictors Mg, Ca, Cr, Al, and Pb were highly significant (p < 0.01), Se, Cl, and NO₂ were significant (p < 0.05), and PO₄ and SO₄ were significant (p < 0.1). The model showed R-squared (adj) 0.85 and deviance explained 88.1%. PMF identified four main emission sources of PM₁₀ in Makkah: (1) Road traffic emissions (explained 51% variance); (2) Industrial emissions and mineral dust (explained 27.5% variance); (3) Restaurant and dwelling emissions (explained 13.6% variance); and (4) Fossil fuel combustion (explained 7.9% variance). Full article