Influence diagnostics in mixed effects logistic regression models
Correlated binary responses are commonly described by mixed effects logistic regression models. This article derives a diagnostic methodology based on the Q-displacement function to investigate local influence of the responses in the maximum likelihood estimates of the parameters and in the predictive performance of the mixed effects logistic regression model. An appropriate perturbation strategy of the probability of success is established, as a form of assessing the perturbation in the response. The diagnostic methodology is evaluated with Monte Carlo simulations. Illustrations with two real-world data sets (balanced and unbalanced) are conducted to show the potential of the proposed methodology.
KeywordsApproximation of integrals Correlated binary responses Metropolis–Hastings and Monte Carlo methods Probability of success R software
Mathematics Subject Classification62J20 62J12
The authors thank the Editors and two referees for their constructive comments on an earlier version of this manuscript which resulted in this improved version. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001, by HPC resources provided by the Information Technology Superintendence of the University of São Paulo, and also by CNPq from Brazil; as well as by the Chilean Council for Scientific and Technology Research (CONICYT) through fellowship “Becas-Chile” (A. Tapia) and FONDECYT 1160868 Grant (V. Leiva) from the Chilean government.
- Crowder MJ (1978) Beta-binomial ANOVA for proportions. J R Stat Soc C 27:34–37Google Scholar
- Hossain M, Islam MA (2003) Application of local influence diagnostics to the linear logistic regression models. Dhaka Univ J Sci 51:269–278Google Scholar
- R Core Team (2016) R: a language and environment for statistical computing. R Foundation for Statistical Computing, ViennaGoogle Scholar