{"id":386,"date":"2017-05-12T11:18:49","date_gmt":"2017-05-12T11:18:49","guid":{"rendered":"http:\/\/jsr.isrt.ac.bd\/?post_type=article&p=386"},"modified":"2017-05-12T11:19:21","modified_gmt":"2017-05-12T11:19:21","slug":"variational-bayesian-logistic-regression-model-selection-improvement-laplace","status":"publish","type":"article","link":"http:\/\/jsr.isrt.ac.bd\/article\/variational-bayesian-logistic-regression-model-selection-improvement-laplace\/","title":{"rendered":"Variational Bayesian logistic regression model selection: an improvement over Laplace?"},"content":{"rendered":"
Increasingly, statisticians are faced with the problem of identifying interesting
\nsubsets of predictors from among a large number of candidates. Existing methods
\nfor variable selection, such as stochastic search algorithms, tend to explore
\nthe model space too slowly in large dimensions. Shotgun stochastic search (SSS)
\nalgorithms have been proposed as an efficient alternative. As current SSS algorithms
\nrely on conjugacy, they are not appropriate for generalized linear models
\nwithout use of approximation methods. This article compares the frequently used
\nLaplace approximation with two alternatives based on Variational Bayes methods.
\nThe comparison is illustrated using several simulated data examples and
\nan application to the problem of predicting conception using data on timing of
\nintercourse in the menstrual cycle. This application also illustrates the problem
\nof selection of interactions.<\/p>\n