Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap
Abstract
1. Introduction
Literature Review
2. Materials and Methods
2.1. Study Design and Data Source
Variable Overview and Analysis Framework
2.2. Statistical Framework and Methodological Rational
- 1.
- Partial Tests:
- 2.
- Combination Functions:
- 3.
- Multiple Testing Correction:
- 4.
- NPC Ranking:
2.2.1. Rationale for Bootstrap Resampling
- Distribution-Free Estimation: It avoids assuming normality or specific parametric distributions, which is particularly beneficial in clinical datasets with skewed, heavy-tailed, or censored values.
- Robustness in Small Samples: The empirical nature of bootstrap confidence intervals provides more reliable estimation of variability when traditional asymptotic theory fails due to limited data.
2.2.2. Exactness of Permutation-Based NPC Testing
2.3. Bootstrap Resampling for Estimation
- 1000 resamples with replacement from the original data.
- Estimation of means, bias, standard error, and percentile-based 95% confidence intervals.
- Implementation using the boot package in R version 4.3.1.
2.4. Software and Reproducibility
- boot for bootstrap analysis
- dplyr and tidyverse for data manipulation
- pheatmap for correlation matrix visualization
2.5. Ethical Consideration
2.6. Use of Generative AI
3. Results
3.1. Partial and Global Hypothesis Testing
3.1.1. Partial Test Results
3.1.2. Global NPC Tests via Combination Functions
- HBA1C alone could drive the significance in Tippett’s test.
- Fisher’s method shows joint significance.
- Lipták’s method balances the contribution of each test, reaffirming clinical relevance.
3.1.3. Bonferroni Adjustment
- Adjusted p-value for HBA1C: ≈ 2.08 × 10−154
- Adjusted p-value for ALBUMINA: ≈ 0.534
3.2. NPC Global Ranking of Patient Severity
3.2.1. Construction of Severity Rankings
- Individual Variable Ranking: Each patient was separately ranked according to their values for each variable—HBA1C, ALBUMINA, MORTO (mortality), and AMPUTAZIONE (major amputation) [7]. For continuous variables (HBA1C, ALBUMINA), lower values—indicating poorer metabolic or nutritional profiles—were assigned lower (worse) ranks. For binary outcomes, deceased or amputated cases received the lowest ranks.
- Global Score Calculation: The individual ranks across all four variables were summed for each patient to compute a composite global severity score. This summation approach ensures that patients with consistently poor profiles across several dimensions are recognized as more severe, whereas those with discrepant or borderline values receive intermediate scores.
- Sorting and Identification of Highest-Risk Profiles: Patients were then sorted according to their global scores (with lower summed scores indicating higher severity). This enabled both a ranked list of clinical severity and the identification of patients at the highest risk for targeted intervention.
3.2.2. Visualization of Ranking Distribution
3.3. Bootstrap Estimations for Clinical Biomarkers
3.3.1. HBA1C Resampling Results
- Mean: 0.0894
- Bias: ≈ 0.0001
- Standard Error: 0.002
- 95% Confidence Interval: (0.0830, 0.0957)
- Bootstrap confirms HBA1C levels are consistently and precisely below clinical thresholds (see Table 7);
- Minimal bias and low SE indicate highly stable estimates.
3.3.2. ALBUMINA Resampling Results
- Mean: 2.723
- Bias: ≈ 0.006
- Standard Error: 0.087
- 95% Confidence Interval: (2.551, 2.866)
- The mean is near the clinical cutoff (2.8 g/dL) (see Table 7);
- Higher SE reflects biological and nutritional variability.
3.3.3. Visualization of Variable Patterns
3.4. Correlation Analysis
3.5. Mathematical Formulations of NPC Methods
- Define the Partial Hypotheses
- 2.
- Formulate the Global Hypothesis
- 3.
- Select Suitable Test Statistics
- 4.
- Permute the Data
- 5.
- Compute Partial p-values
- 6.
- Combine Partial p-values Use a combining function to synthesize the partial p-values into a single global test statistic. Common combining functions include:
- Fisher’s Combination Method
- b.
- Tippett’s Combination Method
- c.
- Lipták’s Combination Method
- 7.
- Evaluate the Global Test Statistic
- Rare disease modeling
- Genomic and gene expression studies
- Subgroup analysis in clinical trials
- Multi-endpoint or composite-outcome evaluation
- Strong Family-Wise Error Rate (FWER) Control
- 2.
- Flexibility
4. Discussion
4.1. Comparison with Established Clinical Scoring Systems
4.2. Clinical Interpretation of Biomarker Patterns
4.3. Insights from Recent Comparative Methodology
4.4. Translational Potential: From Research to Real-Time Decision Support
Limitations and Future Work
- Max-type combination functions for detecting extreme deviations,
- Inverse normal transforms with clinically informed weights [28],
- Adaptive weighting schemes based on variable effect sizes, clinical importance, or prior knowledge.
5. Conclusions
Author Contributions
Funding
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
Abbreviations
Abbreviation | Full Form |
NPC | Non-Parametric Combination |
CI | Confidence Interval |
FWER | Family-Wise Error Rate |
HBA1C | Glycated Hemoglobin |
ALB | Albumin |
IRB | Institutional Review Board |
RANK_GLOBAL | Global Severity Rank (NPC Aggregate) |
SE | Standard Error |
Appendix A
Supplementary Mathematical Details
Variable | Mean | Bias | SE | 95% CI |
---|---|---|---|---|
HBA1C | 0.0894 | 0.0001 | 0.002 | (0.0830, 0.0957) |
ALBUMINA | 2.723 | 0.006 | 0.087 | (2.551, 2.866) |
Appendix B. R Code Snippets and Workflow
- Code B4. NPC Statistical Tests and Bonferroni Correction
- # Partial p-values from one-sample t-tests
p_values <- c( HBA1C = tryCatch(t.test(data$HBA1C, mu = 7)$p.value, error = function(e) NA), ALBUMINA = tryCatch(t.test(data$ALBUMINA, mu = 2.8)$p.value, error = function(e) NA) )
-
# Fisher's combination of p-values
combined_stat <- -2 * sum(log(p_values), na.rm = TRUE) df <- 2 * sum(!is.na(p_values)) p_fisher <- 1 - pchisq(combined_stat, df)
-
# Bonferroni correction
p_bonf <- p.adjust(p_values, method = "bonferroni")
-
# Output
cat("\nPartial p-values:\n"); print(p_values) cat("\nFisher Combined p-value:", round(p_fisher, 5), "\n") cat("\nBonferroni Corrected p-values:\n"); print(p_bonf)
-
Code B5. Descriptive Statistics for Table 3
-
# Descriptive statistics for HBA1C and ALBUMINA
descriptive_stats <- data %>% summarise( HBA1C_MEAN = mean(HBA1C), HBA1C_MEDIAN = median(HBA1C), HBA1C_SD = sd(HBA1C), HBA1C_MIN = min(HBA1C), HBA1C_MAX = max(HBA1C), ALBUMINA_MEAN = mean(ALBUMINA), ALBUMINA_MEDIAN = median(ALBUMINA), ALBUMINA_SD = sd(ALBUMINA), ALBUMINA_MIN = min(ALBUMINA), ALBUMINA_MAX = max(ALBUMINA) )
-
# Prepare formatted table
table3 <- data.frame( Variable = c("HBA1C", "ALBUMINA"), Mean = c(descriptive_stats$HBA1C_MEAN, descriptive_stats$ALBUMINA_MEAN), Median = c(descriptive_stats$HBA1C_MEDIAN, descriptive_stats$ALBUMINA_MEDIAN), SD = c(descriptive_stats$HBA1C_SD, descriptive_stats$ALBUMINA_SD), Min = c(descriptive_stats$HBA1C_MIN, descriptive_stats$ALBUMINA_MIN), Max = c(descriptive_stats$HBA1C_MAX, descriptive_stats$ALBUMINA_MAX) ) cat("\nDescriptive Statistics:\n"); print(table3, row.names = FALSE)
-
Code B6. Bootstrap Confidence Intervals
-
# Bootstrap function
boot_func <- function(x, indices) mean(x[indices]) # Apply bootstrap for HBA1C and ALBUMINA boot_results <- list( HBA1C = boot(data$HBA1C, statistic = boot_func, R = 1000), ALBUMINA = boot(data$ALBUMINA, statistic = boot_func, R = 1000) )
-
# Bootstrap percentile-based confidence intervals
cat("\nBootstrap CI for HBA1C:\n") print(boot.ci(boot_results$HBA1C, type = "perc")) cat("\nBootstrap CI for ALBUMINA:\n") print(boot.ci(boot_results$ALBUMINA, type = "perc"))
-
Code B7. Global NPC Ranking of Patients
-
# Rank individual variables and compute global severity rank
ranked <- data %>% mutate(across(everything(), ~ rank(., ties.method = "average"), .names = "RANK_{.col}")) %>% mutate(RANK_GLOBALE = rowSums(select(., starts_with("RANK_")))) %>% arrange(RANK_GLOBALE)
-
# Top 10 most severe profiles
cat("\nTop 10 Most Severe Patients:\n"); print(head(ranked, 10))
References
- Anaya, D.A.; Dellinger, E.P. Necrotizing soft-tissue infection: Diagnosis and management. Clin. Infect. Dis. 2007, 44, 705–710. [Google Scholar] [CrossRef] [PubMed]
- Raveendranadh, A.; Prasad, S.S.; Viswanath, V. Necrotizing fasciitis: Treatment concepts & clinical outcomes—An institutional experience. BMC Surg. 2024, 24, 336. [Google Scholar]
- Suraphee, S.; Busababodhin, P.; Chamchong, R.; Suparatanachatpun, P.; Khamthong, K. Modified Laboratory Risk Indicator and Machine Learning in classifying necrotizing fasciitis from cellulitis patients. Appl. Sci. 2024, 14, 9241. [Google Scholar] [CrossRef]
- Kim, J.; Yoo, G.; Lee, T.; Kim, J.H.; Seo, D.M.; Kim, J. Classification model for diabetic foot, necrotizing fasciitis, and osteomyelitis. Biology 2022, 11, 1310. [Google Scholar] [CrossRef]
- Cai, Y.; Cai, Y.; Shi, W.; Feng, Q.; Zhu, L. Necrotizing fasciitis of the breast: A review of the literature. Surg. Infect. 2021, 22, 363–373. [Google Scholar] [CrossRef] [PubMed]
- Roje, Z.; Roje, Z.; Matić, D.; Librenjak, D.; Dokuzović, S.; Varvodić, J. Necrotizing fasciitis: Literature review of contemporary strategies for diagnosing and management with three case reports: Torso, abdominal wall, upper and lower limbs. World J. Emerg. Surg. 2011, 6, 46. [Google Scholar] [CrossRef] [PubMed]
- Healy, M.A.; Davis, K.A.; Pepple, C.; Claridge, J.A. Risk factors for mortality in patients with necrotizing soft-tissue infections. Surgery 2016, 160, 1279–1288. [Google Scholar]
- Stevens, D.L.; Bisno, A.L.; Chambers, H.F.; Dellinger, E.P.; Goldstein, E.J.; Gorbach, S.L.; Hirschmann, J.V.; Kaplan, S.L.; Montoya, J.G.; Wade, J.C. Practice guidelines for the diagnosis and management of skin and soft tissue infections. Clin. Infect. Dis. 2014, 59, e10–e52. [Google Scholar] [CrossRef]
- Altman, D.G.; Bland, J.M. Statistics notes: The normal distribution. BMJ 1995, 310, 298. [Google Scholar] [CrossRef]
- Giacalone, M.; Alibrandi, A. Overview and main advances in permutation tests for linear regression models. J. Math. Syst. Sci. 2015, 5, 53–59. [Google Scholar] [CrossRef]
- Goeman, J.J.; Solari, A. Multiple testing for exploratory research. Stat. Sci. 2014, 29, 584–597. [Google Scholar] [CrossRef]
- Pesarin, F. Multivariate Permutation Tests: With Applications in Biostatistics; Wiley: Hoboken, NJ, USA, 2001. [Google Scholar]
- Bonnini, S.; Assegie, G.M.; Kamila, T. Review about the permutation approach in hypothesis testing. Mathematics 2024, 12, 2617. [Google Scholar] [CrossRef]
- Pesarin, F.; Salmaso, L. Permutation Tests for Complex Data: Theory, Applications and Software; Wiley: Hoboken, NJ, USA, 2010. [Google Scholar]
- Bonnini, S.; Corain, L.; Pesarin, F.; Salmaso, L. Nonparametric Statistics for Complex Data: Methods and Case Studies; Wiley: Hoboken, NJ, USA, 2014. [Google Scholar]
- Hemerik, J.; Thoresen, M.; Finos, L. Permutation testing in high-dimensional linear models: An empirical investigation. J. Stat. Comput. Simul. 2021, 91, 897–914. [Google Scholar] [CrossRef]
- Stratton, I.M.; Adler, A.I.; Neil, H.A.W.; Matthews, D.R.; Manley, S.E.; Cull, C.A.; Hadden, D.; Turner, R.C.; Holman, R.R. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): Prospective observational study. BMJ 2000, 321, 405–412. [Google Scholar] [CrossRef] [PubMed]
- Vincent, J.L.; Dubois, M.J.; Navickis, R.J.; Wilkes, M.M. Hypoalbuminemia in acute illness: Is there a rationale for intervention? Ann. Surg. 2003, 237, 319–334. [Google Scholar] [CrossRef]
- Dhatariya, K.; Levy, N.; Kilvert, A.; Watson, B.; Cousins, D.; Flanagan, D.; Hilton, L.; Sinclair, A.; Rayman, G. NHS Diabetes guideline for the perioperative management of the adult patient with diabetes. Diabet. Med. 2012, 29, 420–433. [Google Scholar] [CrossRef]
- Fisher, R.A. Statistical Methods for Research Workers, 4th ed.; Oliver and Boyd: Edinburgh, UK, 1932. [Google Scholar]
- Tippett, L.H.C. The Methods of Statistics; Williams & Norgate: London, UK, 1931. [Google Scholar]
- Lipták, T. On the combination of independent tests. Magy. Tudomanyos Akad. Mat. Kut. Intézetének Közleményei 1958, 3, 171–197. [Google Scholar]
- Bonferroni, C.E. Teoria statistica delle classi e calcolo delle probabilità. Pubbl. R Ist. Super. Sci. Econ. Commer. Firenze 1936, 8, 3–62. [Google Scholar]
- Giacalone, M.; Zirilli, A.; Alibrandi, A.; Cozzucoli, P.C. Bonferroni-Holm and permutation tests to compare health data: Methodological and applicative issues. BMC Med. Res. Methodol. 2018, 18, 81. [Google Scholar] [CrossRef] [PubMed]
- Wang, Y.; Li, C.; Yin, G.; Wang, J.; Li, J.; Wang, P.; Bian, J. Extraction parameter optimized radiomics for neoadjuvant chemotherapy response prognosis in advanced nasopharyngeal carcinoma. Clin. Transl. Radiat. Oncol. 2022, 33, 37–44. [Google Scholar] [CrossRef]
- Alibrandi, A.; Giacalone, M.; Zirilli, A. Psychological stress in nurses assisting Amyotrophic Lateral Sclerosis patients: A statistical analysis based on Non-Parametric Combination test. Mediterr. J. Clin. Psychol. 2022, 10. [Google Scholar] [CrossRef]
- Andreella, A.; Hemerik, J.; Finos, L.; Weeda, W.; Goeman, J. Permutation-based true discovery proportions for functional magnetic resonance imaging cluster analysis. Stat. Med. 2023, 42, 2311–2340. [Google Scholar] [CrossRef] [PubMed]
- Bonnini, S.; Borghesi, M.; Giacalone, M. Advances on multisample permutation tests for “V-shaped” and “U-shaped” alternatives with application to circular economy. Ann. Oper. Res. 2024, 342, 1655–1670. [Google Scholar] [CrossRef]
- Hoesl, V.; Kempa, S.; Prantl, L.; Ochsenbauer, K.; Hoesl, J.; Kehrer, A.; Bosselmann, T. The LRINEC Score—An indicator for the course and prognosis of necrotizing fasciitis? J. Clin. Med. 2022, 11, 3583. [Google Scholar] [CrossRef] [PubMed]
- Johnson, S.R.; Benvenuti, T.; Nian, H.; Thomson, I.P.; Baldwin, K.; Obremskey, W.T.; Moore-Lotridge, S.N. Measures of admission immunocoagulopathy as an indicator for in-hospital mortality in patients with necrotizing fasciitis: A retrospective study. JBJS Open Access 2023, 8, e22. [Google Scholar] [CrossRef]
- Bonnini, S.; Borghesi, M.; Giacalone, M. Simultaneous marginal homogeneity versus directional alternatives for multivariate binary data with application to circular economy assessments. Appl. Stoch. Models Bus. Ind. 2024, 40, 389–407. [Google Scholar] [CrossRef]
- Pesarin, F.; Salmaso, L. The permutation testing approach: A review. Statistica 2010, 70, 481–509. [Google Scholar]
- Suraphee, S.; Sukmun, C.; Khamthong, K.; Wichitchan, S.; Chamchong, R. Classification of necrotizing fasciitis from cellulitis patients using machine learning in Maha Sarakham Hospital, Thailand. In Proceedings of the 2024 International Conference on Machine Learning and Cybernetics (ICMLC), Miyazaki, Japan, 20–23 September 2024; IEEE: New York, NY, USA, 2024; pp. 553–558. [Google Scholar]
- Giacalone, M.; Zirilli, A.; Moleti, M.; Alibrandi, A. Does the iodized salt therapy of pregnant mothers increase the children’s IQ? Empirical evidence of a statistical study based on permutation tests. Qual. Quant. 2018, 52, 1423–1435. [Google Scholar] [CrossRef]
- Manandhar, K.; Agrawal, S. SIARI Score versus LRINEC Score for diagnosis of necrotizing fasciitis—A prospective comparative study. Res. Artic.-Open Access 2023, 6, 1034. [Google Scholar]
- Arboretti, R.; Barzizza, E.; Ceccato, R. Finding exclusive peak with permutation testing. Comput. Stat. 2025, 1–16. [Google Scholar] [CrossRef]
- Wies, C.; Miltenberger, R.; Grieser, G.; Jahn-Eimermacher, A. Exploring the variable importance in random forests under correlated features. BMC Med. Res. Methodol. 2023, 23, 2023. [Google Scholar] [CrossRef]
- Ferraccioli, F.; Sangalli, L.M.; Finos, L. Nonparametric tests for semiparametric regression models. Test 2023, 32, 1106–1130. [Google Scholar] [CrossRef]
- Pesarin, F.; Salmaso, L. A review and some new results on permutation testing for multivariate problems. Stat. Comput. 2012, 22, 639–646. [Google Scholar] [CrossRef]
- Corain, L.; Salmaso, L. Improving power of multivariate combination-based permutation tests. Stat. Comput. 2015, 25, 203–214. [Google Scholar] [CrossRef]
- Finos, L.; Pesarin, F. On zero-inflated permutation testing and some related problems. Stat. Pap. 2020, 61, 2157–2174. [Google Scholar] [CrossRef]
- Orlenko, A.; Moore, J.H. A comparison of methods for interpreting random forest models of genetic association in the presence of non-additive interactions. BioData Min. 2021, 14, 9. [Google Scholar] [CrossRef]
Traditional Methods | Limitations | Proposed Solution |
---|---|---|
Logistic Regression | Assumes normality, large n | NPC avoids distributional assumptions |
ANOVA | Poor with mixed variable types | NPC handles heterogeneity |
Classical p-values | Inflated error in multiple testing | Combination functions (Fisher, Tippett) |
Aspect | Advantages | Disadvantages |
---|---|---|
Assumptions | Less rigid requirements; do not need data to be normally distributed or homoscedastic | May be less powerful if data actually follow parametric assumptions |
Data Types | Applicable to diverse data: ranks, ordered, categorical, binary (e.g., gender, yes/no) | Sometimes require reduction of continuous data to ranks/order, leading to loss of information |
Flexibility | Suitable for small samples, heterogeneous or mixed-type variables | Results can be less efficient: typically require stronger evidence (larger differences, samples) |
Robustness | Do not rely on model structure; robust to outliers and data irregularities | Interpretation may be less familiar to clinicians used to classic effect sizes or odds ratios |
Variable | Type | Clinical Meaning |
---|---|---|
HBA1C | Continuous | Marker of long-term glycemic control; elevated or abnormally low values may reflect metabolic dysregulation. |
ALBUMINA | Continuous | Indicator of nutritional status and systemic inflammation; low values often signal poor prognosis in critical illness. |
MORTO | Binary | Mortality status; coded as 0 (survived) and 1 (deceased). |
AMPUTAZIONE | Binary | Surgical outcome; coded as 0 (no amputation) and 1 (major amputation). |
Variable | Test Type | Reference Value | p-Value | Interpretation |
---|---|---|---|---|
HBA1C | One-sample t-test | 7.0 | 1.04 × 10−154 | Significant deviation (very low) |
ALBUMINA | One-sample t-test | 2.8 | 0.267 | Not statistically significant |
Combination Method | Global p-Value | Interpretation |
---|---|---|
Fisher | 0.000 | Strong joint significance across variables |
Tippett | 1.04 × 10−154 | Dominance of HBA1C’s extreme significance |
Lipták | <0.001 | Balanced evidence across tests |
S.No | AMPUTAZIONE | HBA1C | MORTO | ALBUMINA | RANK_ | RANK | RANK_ | RANK_ | RANK_ |
---|---|---|---|---|---|---|---|---|---|
AMPUTAZIONE | _HBA1C | MORTO | ALBUMINA | GLOBALE | |||||
1 | 0 | 0.048 | 0 | 1.9 | 21.5 | 1 | 23.5 | 9 | 55 |
2 | 0 | 0.058 | 0 | 1.84 | 21.5 | 7 | 23.5 | 5 | 57 |
3 | 0 | 0.058 | 0 | 1.87 | 21.5 | 7 | 23.5 | 7 | 59 |
4 | 0 | 0.057 | 0 | 2.4 | 21.5 | 5 | 23.5 | 21 | 71 |
5 | 0 | 0.078 | 0 | 1.9 | 21.5 | 23.5 | 23.5 | 9 | 77.5 |
6 | 0 | 0.059 | 0 | 2.5 | 21.5 | 9 | 23.5 | 27 | 81 |
7 | 0 | 0.060 | 0 | 2.5 | 21.5 | 11 | 23.5 | 27 | 83 |
8 | 0 | 0.072 | 0 | 2.3 | 21.5 | 22 | 23.5 | 16.5 | 83.5 |
9 | 0 | 0.054 | 0 | 2.74 | 21.5 | 2 | 23.5 | 38 | 85 |
10 | 0 | 0.055 | 0 | 2.8 | 21.5 | 3 | 23.5 | 40 | 88 |
Variable | Mean | Bias | SE | 95% CI | Interpretation |
---|---|---|---|---|---|
HBA1C | 0.0894 | 0.0001 | 0.002 | (0.0830, 0.0957) | Low bias, highly precise |
ALBUMINA | 2.723 | 0.006 | 0.087 | (2.551, 2.866) | Greater variability observed |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Piscopo, G.; Bandaru, S.T.; Giacalone, M.; Longobardi, M. Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap. Mathematics 2025, 13, 2869. https://doi.org/10.3390/math13172869
Piscopo G, Bandaru ST, Giacalone M, Longobardi M. Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap. Mathematics. 2025; 13(17):2869. https://doi.org/10.3390/math13172869
Chicago/Turabian StylePiscopo, Gianfranco, Sai Teja Bandaru, Massimiliano Giacalone, and Maria Longobardi. 2025. "Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap" Mathematics 13, no. 17: 2869. https://doi.org/10.3390/math13172869
APA StylePiscopo, G., Bandaru, S. T., Giacalone, M., & Longobardi, M. (2025). Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap. Mathematics, 13(17), 2869. https://doi.org/10.3390/math13172869