Dear Alex, You have just rediscovered one of the problems of classical testing - when the sample size is large, you have the power to detect differences that are statistically significant, but have no practical importance. The answer here is to filter on a threshold of differential expression which you consider to be important in addition to statistical significance (2-fold is often used) or switch to Bayesian methods (e.g. use the Bayes factor). --Naomi At 02:40 PM 9/11/2007, Alex Tsoi wrote: >Dear all, > >Currently I am using SAM to identify the differentially expressed genes >between two groups. The first group has 99 experiments, and the second one >has around 15 experiments. When I use the SAM to do the analysis, I identify >so many differentially expressed genes even if I set the FDR threshold to be >very low. I am just wondering if I should just randomly pick 15 experiments >from the first group, and to compare the 15 experiments on the second group >? or does it matter ? > >I greatly appreciate for any comment or advice. > > >Thanks, > > > > >-- >Lam C. Tsoi (Alex) >Medical University of South Carolina > > [[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