Bayesian logistic regression and its application for hypothyroid prediction in post-radiation nasopharyngeal cancer patients

Abstract

Logistic regression models are commonly used to model response variables in the form of categorical variables with several predictor variables. The contribution of the predictor variable to the response variable is expressed through a regression coefficient (β). Therefore, it is necessary to estimate β. This study discusses the estimation of β using the Bayesian method. Bayesian approach utilizes a combination of information from sample data and prior information about the characteristics of the parameters of interest, resulting in the updated information, namely the posterior. Bayesian method thus can overcome the problem if the quality of the sample data does not support observation. Bayesian logistic regression method will be used in analyzing post-radiation nasopharyngeal cancer (NPC) patient data, using measurement on Zulewski's score components. Markov Chain Monte Carlo with Gibbs Sampling were used to obtain the sample from posterior distribution. Convergent estimates were obtained, and the result showed that Zulewski's component scores only were not enough to explain the hypothyroidism in NPC. Additional information is required in order to explain the incidence of hypothyroidism in NPC.