Validating negative binomial lyme disease regression model with bootstrap resampling. (August 2016)

• Record Type: Journal Article
• Authors: Tran, Phoebe; Tran, Lam
• Is Part Of: Environmental modelling & software. Volume 82(2016:Aug.)
• Page Start: 121
• Page End: 127
• Publication Date: 2016-08
• Subjects: Lyme disease -- Bootstrap resampling -- Landscape fragmentation -- Bootstrap confidence intervals; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2016.04.019
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Various negative binomial regression models have been developed to study Lyme disease in connection to climate and/or landscape factors. However, no internal validation of any of those models has been reported in the literature. This study used bootstrap resampling to conduct an internal validation of a negative binomial regression model on Lyme disease incidence. The model used county-level Lyme disease incidence in thirteen states in the Northeastern United States during 2002–2006 and linked it with several previously identified key landscape and climatic variables used in an earlier study. Results showed that there were significant differences between the outcomes from the initial model and those from bootstrap resampling. Arguably bootstrap resampling, as illustrated in this study, can serve as a sound and valuable means to provide a second line of evidence on model outcomes and shed more insight on variables (e.g., climate and landscape factors) included in the models. Highlights: Negative binomial Lyme disease model was validated via bootstrap resampling. Significant difference between initial model's outcomes and bootstrapping results. Bootstrapping results might be inconsistent due to data dependence/spatial heterogeneity. Bootstrap is a sound method for a second line of evidence of model outcomes.