Bayesian Variable Selection in Markov Mixture Models |
| |
| Authors: | Roberta Paroli Luigi Spezia |
| |
| Affiliation: | 1. Dipartimento di Scienze Statistiche , Università Cattolica S.C. , Milano, Italy roberta.paroli@unicatt.it;3. Dipartimento di Statistica , Università Ca' Foscari , Venezia, Italy |
| |
| Abstract: | Monte Carlo simulation is used to evaluate the actual confidence levels of five different approximations for confidence intervals for the probability of success in Markov dependent trials. The approximations involve the conditional probability of success as a nuisance parameter, and the effects of substituting Klotz's (1973), Price's (1976), and a new estimator are also evaluated. The new estimator is less biased and tends to increase the confidence level. A program for calculating the estimator and the confidence interval approximations is available. |
| |
| Keywords: | Gibbs variable selection Kuo–Mallick method Metropolized–Kuo–Mallick method Stochastic search variable selection |
|
|
