Open AccessA BAYESIAN APPROACH TO THE SELECTION OF PREDICTOR VARIABLES
Author(s) -
Novick Melvin R.
Publication year - 1969
Publication title -
ets research bulletin series
Language(s) - English
Resource type - Journals
eISSN - 2333-8504
pISSN - 0424-6144
DOI - 10.1002/j.2333-8504.1969.tb00736.x
Subject(s) - bayesian probability , selection (genetic algorithm) , statistics , regression , bayesian linear regression , regression analysis , econometrics , model selection , correlation , mathematics , feature selection , bayesian inference , computer science , artificial intelligence , geometry
Some work of Lindley on a Bayesian structural model is shown to be relevant to the problem of the selection of predictor variables. It is suggested that estimates of regression parameters be regressed toward a mean value and that the resulting attenuated estimate of the multiple correlation be adopted. These regressed estimates are derived from a general Bayesian normal law analysis.