Results will be returned via a link provided by e-mail. The results will be available on the site for at least one week.

Data Format

DATA: A matrix containing the data to be analyzed. Each row of the matrix should contain one observation. The first column of each row should hold the output, Y, while the remaining columns (2 to num_vars +1, where num_vars is the total number of variables) should correspond to each of the predictors. i.e. DATA = [Y X].

The number of rows must be larger than the number of columns and the number of columns cannot exceed 30.
num_results: i. (Default) If num_results = 0, then EBIR will return the posterior probability for every possible sub-model.

ii. If a value for num_results > 0 is specified by the user, then EBIR will only return the best num_results sub-models.

k_i, k_e: The scale parameters for the multivariate normal prior distribution on the regression coefficients. As a general rule, the inclusion parameter should be chosen so that 0 < k_i < 1, while the exclusion parameter should be large, k_e > 1. If no values are entered by the user, the default parameter settings are: k_i =0.01, k_e = 100.

p_i: The prior probability of including a variable in the regression. Note: p_e is the exclusion probability [p_e = 1-p_i]. By default, all sub-models are equally likely, i.e. p_i = p_e =0.5.

v_0, sig_0: The parameters for the Inverse chi square prior distribution on the error variance. These parameters act as v_0 pseudo data points of variance sig_0. By default, v_0=5 and sig_0 =1.

Last modified: Thu Oct 27 16:31:06 EDT 2011