How To Estimate The False Discovery Rate (Fdr) Using Bootstrap

Question

[YJ]

Hi How can I estimate the false discovery rate (FDR) using bootstrap in R, say, 3000 simple linear regression tests with cut-off = 6. Thanks

[David W]

Going to need a bit more. First, what's the bioinformatics angle. Second, your brief description of FDR seems a bit confused, do you want to spell out exactly what data you have, and what you like to do with it?

[Neilfws]

Yes, much more detail required to answer this question.

[Raygozak]

I think this question is better suited for "crossvalidated" (http://stats.stackexchange.com/) in stack exchange.

Answer 1

[David]

Bootstrap is a method to for obtaining standard errors and CI when you want to improve the one your procedure give you or computing them on your real dataset is intractable or not available.

You can cannot compute FDR using bootstrap. FDR can be computed using a re-sampling approach without replacement which is different from the bootstrap approach (with replacement). I wrote a how-to on my blog to compute pFDR in R.