National Cancer Institute Home at the National Institutes of Health | www.cancer.gov

Publication Search Results - Publication Detail

Title: Bayesian model selection for join point regression with application to age-adjusted cancer rates
Authors: Tiwari RC,  Cronin KA,  Davis WW,  Feuer EJ,  Yu B,  Chib S
Journal: Journal of the Royal Statistical Society, Series C, Applied Statistics
Date: 2005 Nov
PubMed ID:
PMC ID: not available
Abstract: A Bayesian analysis of joinpoint regression model is considered and a model selection approach based on the Bayes factor is developed. Given a set of (K+1) joinpoint models, with 0,1,..., K, joinpoints, the method compares any two joinpoint models through the Bayes factor (BF) that requires computation of the marginal likelihood of each competing model from the output of Markov chain Monte Carlo simulations. The Bayesian model selection of joinpoint regression is encompassed by the prior-posterior analysis of the parameters of the joinpoint model. The model with the largest BF is selected as the best model. Another approach based on Bayes information criterion (BIC) is also developed. The model with the smallest value of the BIC is selected as the best model. The method is applied to analyze the observed U.S. cancer incidence rates for some selected cancer sites from the National Cancer Institute's SEER program. The graphs of the joinpoint models to the data are produced using BF, BIC and the permutation test based (PTB) method (by Kim et al. (2000), Statistics in Medicine, vol. 19, 335-351). The Bayesian joinpoint model and model selection method presented here will be integrated in the National Cancer Institute's PTB Joinpoint software available at: http://www.srab.cancer.gov/joinpoint/.