Thank you Sean. Now I am clear how limma handle multiple correction with multiple correction. It's using "separate" mode. decideTests() source code clearly show this. Thank you. Regards, Xiaokuan ________________________________ From: Sean Davis <sdavis2@mail.nih.gov> Cc: bioconductor <bioconductor@stat.math.ethz.ch> Sent: Mon, September 27, 2010 1:39:11 PM Subject: Re: [BioC] applying p.adjust() in limma Dear List, > >Sorry to bring up this adjusted p value question about limma. I searched mail >list but didn't find an answer. > >I know limma use "p.adjust()" for multiple test correction. And if there is only >one coefficent, the adjust p values results of limma's top table is the same as >I directly apply p.adjust to the P.Value vector. > >My question is that if we have several coefficent, for example, 3 comparisons, >how does limma use the p.adjust()? >I tried to feed p.adjust() as: p.adjust(tp$P.Value, >method="fdr",n=length(tp$P.Value) * 3), but the results are different from the >tp$adj.P.Val. > >Could some let me know how limma apply p.adjust() to handle this situation >(several comparisons/coefficents)? > > Hi, �Xiaokuan. See the help for decideTests(). Sean [[alternative HTML version deleted]]

Hi Sunny, Sorry for the confusion. In fact, tp is actually one coefficent. i.e. tp <-topTable(fit,coef=1,n=Inf,adj="fdr"). Regards, Xiaokuan ________________________________ From: Sunny Srivastava <research.baba@gmail.com> Cc: bioconductor <bioconductor@stat.math.ethz.ch> Sent: Mon, September 27, 2010 1:46:21 PM Subject: Re: [BioC] applying p.adjust() in limma Hello Wei, If I understand your question correctly, most probably tp$adj.P.Val is the adjusted p value for the moderated F statistic (would be different from the moderated t if the number of treatments is more than 2) When you calculate the p-value separately for the three comparisons and use p.adjust for all of them, this may not be equal to the p values you obtained from F statistic(s). Thanks, S. Dear List, > >Sorry to bring up this adjusted p value question about limma. I searched mail >list but didn't find an answer. > >I know limma use "p.adjust()" for multiple test correction. And if there is only >one coefficent, the adjust p values results of limma's top table is the same as >I directly apply p.adjust to the P.Value vector. > >My question is that if we have several coefficent, for example, 3 comparisons, >how does limma use the p.adjust()? >I tried to feed p.adjust() as: p.adjust(tp$P.Value, >method="fdr",n=length(tp$P.Value) * 3), but the results are different from the >tp$adj.P.Val. > >Could some let me know how limma apply p.adjust() to handle this situation >(several comparisons/coefficents)? > >Thank you. > >Regards, >Xiaokuan > > > >       [[alternative HTML version deleted]] > >_______________________________________________ >Bioconductor mailing list >Bioconductor@stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor >Search the archives: >http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]