Marginal pre-screening algorithm

Given S draws of the p regression coefficients, this function (i) computes (SD-standarized) posterior means, (ii) orders the absolute values of this statistic, and (iii) returns the indices of the largest num_to_keep values.

Usage

prescreen(post_beta, num_to_keep)

Arguments

post_beta

(S x p) matrix of posterior simulations of the p regression coefficients

num_to_keep

number of variables to return in the screening process

Value

ind_keep: the indices of the variables to keep

Details

If the Bayesian model is nonlinear, post_beta may be obtained instead by projecting each vector of posterior predictive draws onto the covariate matrix.