Bayesian inference for logistic models in R language. (Chinese. English summary) Zbl 1374.62096

Summary: In this paper, based on Gibbs sampling and by introducing latent variables of Pólya-gamma distribution families, we make the full conditional distribution of regression coefficient parameters be conjugated normality, and construct Markov chains of the regression coefficient. The estimation of regression coefficient is the posterior mean estimation. Through a set of real data, and introducing R language package of BayesLogit and Glm, we compare the results of the two methods. It is proved that the Pólya-gamma latent variable Bayesian estimation method is effective and precise in dealing with the logistic regression model.


62J12 Generalized linear models (logistic models)
62F15 Bayesian inference


BayesLogit; R
Full Text: DOI