Question

fdr adjustment in limma

0

Entering edit mode

Kimpel, Mark W ▴ 890

@kimpel-mark-w-727

Last seen 9.6 years ago

How does the method "fdr" in limma work? Strictly speaking, I am not aware that fdr methods actually adjust p values but rather set p value cut-offs according the a value q, the user defined acceptable proportion of false positives. The set of all tests with a p value less than or equal to the cut-off would be expected to have a false positive rate of the value q. I have played around with limma's fdr method a bit and I get the feeling that, after fdr correction using p.adjust, perhaps I should set my own p value tolerance equal to my FDR tolerance. In other words, if I am willing to accept an FDR of 10%, after applying p.adjust, I should accept all tests with p<0.10 as significant. Am I right or wrong on this? Thanks, Mark Mark W. Kimpel MD Official Business Address: Department of Psychiatry Indiana University School of Medicine Biotechnology, Research, & Training Center 1345 W. 16th Street Indianapolis, IN 46202 [[alternative HTML version deleted]]

limma limma • 2.4k views

ADD COMMENT • link updated 2.1 years ago by Gordon Smyth 50k • written 19.7 years ago by Kimpel, Mark W ▴ 890

score 0 · Answer 1 · 2004-08-21

0

Entering edit mode

Gordon Smyth 50k

@gordon-smyth

Last seen 2 hours ago

WEHI, Melbourne, Australia

How does the method "fdr" in limma work? Strictly speaking, I am not aware that fdr methods actually adjust p values but rather set p value cut-offs according the a value q, the user defined acceptable proportion of false positives.

Any decision method including fdr can be expressed in terms of adjusted p-values. Please type ?p.adjust and follow the last cited reference.

The set of all tests with a p value less than or equal to the cut-off would be expected to have a false positive rate of the value q.

I have played around with limma's fdr method a bit and I get the feeling that, after fdr correction using p.adjust, perhaps I should set my own p value tolerance equal to my FDR tolerance. In other words, if I am willing to accept an FDR of 10%, after applying p.adjust, I should accept all tests with p<0.10 as significant.

Yes, that's right. That is the way that all p-value adjustment methods work.

Gordon

ADD COMMENT • link 19.7 years ago • updated 2.1 years ago Gordon Smyth 50k

0

Entering edit mode

Hi, I know this has been discussed before: https://stat.ethz.ch/pipermail/bioconductor/2003-July/002025.html but is anyone working on a Bioconductor-style annotation package for the Affy mouse chip 430_2 ? I've downloaded the annotation from the Affy website in a CSV (comma-separated values) file, but it would be nice to have it in a standard Bioconductor format (an R package), especially for affylmGUI. I hoping to avoid coding complicated dialogs in affylmGUI to allow imported arbtirary annotation files, but I'd be happy to hear opinions on whether it would be good to have a simple mechanism to import some CSV annotation files into affylmGUI somehow. I've had a quick look at AnnBuilder, but this seems to be more for adding extra annotation columns from the web, whereas I have plenty of annotation columns already in the CSV file (enough for our purposes). I just want to get them in a standard BioC format. I tried just manually creating an annotation package, but when I tried saving one of the required .rda (RData) files in the data/ directory of the package (e.g. for gene symbols), I found that I then could not load the saved .rda file into R - I got the error Error: protect(): stack overflow I've tried it on my laptop (1GB memory) and on a Linux machine with a lot more memory. So if anyone has an annotation package for this chip that would be great. Or if anyone can steer me in the right direction as far as making an annotation package from the CSV file that would be great. Or if anyone wants to share some R code for reading data from Affy CSV files that would be great too. Regards James

ADD REPLY • link 19.7 years ago James Wettenhall ▴ 1000