An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20080126/ 0d3626dc/attachment.pl

Have a look at the estimate the pi0 - that can have a big effect. To understand what is going on, look at the histograms of unadjusted p-values. The stronger the peak at the low end, the smaller the FDR-adjusted p-values will be. --Naomi At 03:54 AM 1/26/2008, Lana Schaffer wrote: >Hi, >I have a set of samples for which the FDR/adj.Pvalues (BH) >are >.5 for the same unadjusted pvalue level and number of >testing genes as another set of samples. The other set of >samples have more significant FDR/adj.Pvalues. What would >be some of the reasons why these adjusted pvalues would >would be different between the 2 sets of samples? >Lana > > [[alternative HTML version deleted]] > >_______________________________________________ >Bioconductor mailing list >Bioconductor at stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor >Search the archives: >http://news.gmane.org/gmane.science.biology.informatics.conductor Naomi S. Altman 814-865-3791 (voice) Associate Professor Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111

An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20080127/ 87d8bc22/attachment.pl

Since every experiment has false discoveries and false nondiscoveries, you should use time type of confirmatory method to determine if these genes are differentially expressed. --Naomi At 12:30 PM 1/27/2008, Lana Schaffer wrote: >Thanks. >In the experiment where pi0 is probably smaller and the FDR >is high, some of the genes are plausable for the experiment. >In other words, the unadjusted pvalue is small but the FDR is >high for those genes and those genes are relavent to the experiment. >What do you say about those genes? >Lana > >________________________________ > >From: Naomi Altman [mailto:naomi at stat.psu.edu] >Sent: Sat 1/26/2008 6:50 AM >To: Lana Schaffer; bioconductor at stat.math.ethz.ch >Subject: Re: [BioC] FDR/adj.Pvalues > > > >Have a look at the estimate the pi0 - that can have a big effect. To >understand what is going on, look at the histograms of unadjusted >p-values. The stronger the peak at the low end, the smaller the >FDR-adjusted p-values will be. > >--Naomi > >At 03:54 AM 1/26/2008, Lana Schaffer wrote: > >Hi, > >I have a set of samples for which the FDR/adj.Pvalues (BH) > >are >.5 for the same unadjusted pvalue level and number of > >testing genes as another set of samples. The other set of > >samples have more significant FDR/adj.Pvalues. What would > >be some of the reasons why these adjusted pvalues would > >would be different between the 2 sets of samples? > >Lana > > > > [[alternative HTML version deleted]] > > > >_______________________________________________ > >Bioconductor mailing list > >Bioconductor at stat.math.ethz.ch > >https://stat.ethz.ch/mailman/listinfo/bioconductor > >Search the archives: > >http://news.gmane.org/gmane.science.biology.informatics.conductor > >Naomi S. Altman 814-865-3791 (voice) >Associate Professor >Dept. of Statistics 814-863-7114 (fax) >Penn State University 814-865-1348 (Statistics) >University Park, PA 16802-2111 > > > > > [[alternative HTML version deleted]] > >_______________________________________________ >Bioconductor mailing list >Bioconductor at stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor >Search the archives: >http://news.gmane.org/gmane.science.biology.informatics.conductor Naomi S. Altman 814-865-3791 (voice) Associate Professor Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111

BH adjustement (say p.adjust(,...method="BH")) doesn't account for pi0. Moreover, a lower pi0 would give you smaller FDR. So you probably used some othe algorithm. Which one? Stefano On Sun, Jan 27, 2008 at 09:30:57AM -0800, Lana Schaffer wrote: <lana>Thanks. <lana>In the experiment where pi0 is probably smaller and the FDR <lana>is high, some of the genes are plausable for the experiment. <lana>In other words, the unadjusted pvalue is small but the FDR is <lana>high for those genes and those genes are relavent to the experiment. <lana>What do you say about those genes? <lana>Lana <lana> <lana>________________________________ <lana> <lana>From: Naomi Altman [mailto:naomi at stat.psu.edu] <lana>Sent: Sat 1/26/2008 6:50 AM <lana>To: Lana Schaffer; bioconductor at stat.math.ethz.ch <lana>Subject: Re: [BioC] FDR/adj.Pvalues <lana> <lana> <lana> <lana>Have a look at the estimate the pi0 - that can have a big effect. To <lana>understand what is going on, look at the histograms of unadjusted <lana>p-values. The stronger the peak at the low end, the smaller the <lana>FDR-adjusted p-values will be. <lana> <lana>--Naomi <lana> <lana>At 03:54 AM 1/26/2008, Lana Schaffer wrote: <lana>>Hi, <lana>>I have a set of samples for which the FDR/adj.Pvalues (BH) <lana>>are >.5 for the same unadjusted pvalue level and number of <lana>>testing genes as another set of samples. The other set of <lana>>samples have more significant FDR/adj.Pvalues. What would <lana>>be some of the reasons why these adjusted pvalues would <lana>>would be different between the 2 sets of samples? <lana>>Lana <lana>> <lana>> [[alternative HTML version deleted]] <lana>> <lana>>_______________________________________________ <lana>>Bioconductor mailing list <lana>>Bioconductor at stat.math.ethz.ch <lana>>https://stat.ethz.ch/mailman/listinfo/bioconductor <lana>>Search the archives: <lana>>http://news.gmane.org/gmane.science.biology.informatics.conduct or <lana> <lana>Naomi S. Altman 814-865-3791 (voice) <lana>Associate Professor <lana>Dept. of Statistics 814-863-7114 (fax) <lana>Penn State University 814-865-1348 (Statistics) <lana>University Park, PA 16802-2111 <lana> <lana> <lana> <lana> <lana> [[alternative HTML version deleted]] <lana> <lana>_______________________________________________ <lana>Bioconductor mailing list <lana>Bioconductor at stat.math.ethz.ch <lana>https://stat.ethz.ch/mailman/listinfo/bioconductor <lana>Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor