Calculate false discovery rate after analysis

[May Blake]

Hello,

I have run a univariate analysis on multiple variables (800+) in an exploratory analysis. Then I selected significant variables and built a multivariate model with the selected variables. I initially used the Bonferroni to adjust for multiple comparisons but was told using the FDR at 5% will be much better. Can anyone help me with the formula on how to calculate this manually? Is there a way to do this in Stata with my GLM model?

[George Ford]

I wouldn't use that approach. Your X's should be based on theory. You are mining the data in an ad hoc fashion and Bonferroni won't correct for the bias.

If you are just interested in fit, then try lasso.