1. Model selection with multiple regression on distance matrices leads to incorrect inferences
- Author
-
Marie-Josée Fortin, Erin L. Landguth, Ryan P. Franckowiak, Ian S. Acuña-Rodríguez, Karl J. Jarvis, Helene H. Wagner, and Michael Panasci
- Subjects
0106 biological sciences ,0301 basic medicine ,Gene Flow ,Computer and Information Sciences ,Research Validity ,Heredity ,Statistical methods ,Bayesian probability ,lcsh:Medicine ,Statistics (mathematics) ,Research and Analysis Methods ,010603 evolutionary biology ,01 natural sciences ,Models, Biological ,03 medical and health sciences ,Bayesian information criterion ,Statistics ,Geoinformatics ,Genetics ,Statistics::Methodology ,lcsh:Science ,Mathematics ,Evolutionary Biology ,Multidisciplinary ,Population Biology ,Geography ,Model selection ,Simulation and Modeling ,lcsh:R ,Biology and Life Sciences ,Regression analysis ,Random Variables ,Research Assessment ,Probability Theory ,Spatial Autocorrelation ,Deviance information criterion ,Monte Carlo method ,030104 developmental biology ,Sample size determination ,Physical Sciences ,Earth Sciences ,Mathematical and statistical techniques ,Regression Analysis ,lcsh:Q ,Akaike information criterion ,Distance matrices in phylogeny ,Population Genetics ,Research Article - Abstract
In landscape genetics, model selection procedures based on Information Theoretic and Bayesian principles have been used with multiple regression on distance matrices (MRM) to test the relationship between multiple vectors of pairwise genetic, geographic, and environmental distance. Using Monte Carlo simulations, we examined the ability of model selection criteria based on Akaike's information criterion (AIC), its small-sample correction (AICc), and the Bayesian information criterion (BIC) to reliably rank candidate models when applied with MRM while varying the sample size. The results showed a serious problem: all three criteria exhibit a systematic bias toward selecting unnecessarily complex models containing spurious random variables and erroneously suggest a high level of support for the incorrectly ranked best model. These problems effectively increased with increasing sample size. The failure of AIC, AICc, and BIC was likely driven by the inflated sample size and different sum-of-squares partitioned by MRM, and the resulting effect on delta values. Based on these findings, we strongly discourage the continued application of AIC, AICc, and BIC for model selection with MRM.
- Published
- 2017