I'm using Systat software and am trying to find the best predictors (out of about 250) for a categorical predictand (0 or 1). I'm doing spearman and pearson correlations (not sure which is more appropriate) and eliminating the predictors with the lowest R values and eliminating the most highly intercorrelated predictors. Is this a good first step toward achieving some equations that describe the relationship between the best predictors and the predictand/independent variable? I figure I have to use the logit regression functionality eventually but I want to slim down the number of independent variables first. Thanks for any help,
S.