SENSITIVITY OF BAYESIAN DYNAMIC MIXED LOGIT MODELS TO PRIOR DISTRIBUTIONS

Abstract

This research work investigated the performance of various prior distributions on the Bayesian
Dynamic Mixed Logistic Regression Model (BDML). The data set used was a Public datasets
gotten from UCI Machine Learning Repository. The study compared the performance of
Uniform, Jeffrey’s, Exponential, Gamma, Cauchy, Normal, and Beta prior distributions in
capturing the heterogeneity in customer preferences. The result of the Bank marketing data
showed that Jeffery’s prior outperforms other priors used in terms of MAE, RMSE, and Log
Likelihood this showed that the choice of prior distribution significantly affects the model
estimates and predictions.