Quantile normalization reduces the variance among the arrays included. ANOVA compares the between tissue variation to the within tissue variation. It is clear that Method 2 will inflate the false detection rate, by reduces the within tissue variation without adjusting the between tissue variaion. It is not clear that Method 3 sufficiently reduces the problem discussed in 2, although it should "on average" (but not necessarily on a gene by gene basis). I usually use Method 1, although I am sure it inflates the false nondetection rate. I recommend looking at histograms of the PM probes for each array to see if quantile normalization across arrays make sense. In my studies, the variation in the replicate samples was greater than the variation in samples from different tissues, so I felt comfortable using Method 1. --Naomi At 01:50 PM 5/2/2007, shirley zhang wrote: >Dear Bioconductor: > >We run Affy's all exon array on 3 tissues (lung, nose, and mouth). >For each tissue, 10 samples are in control and 10 in treatment group. >Totally 60 samples. > >General speaking, the sigal from lung is similar to that of nose, and >the signal from mouth is much lower than lung and nose. > >I've tried the following normalization: > >Method 1. Quantile Normalize all 3 tissues together >Method 2. Quantile Normalize each tissue separately >Method 3. Quantile Normalize each tissue separately, and then do >Global Median aross all > >My question is: in order to do 2-way ANOVA including the interaction >term between tissue and treatment, which normalization method is >applicable? > >Thanks, > >Shirley > >_______________________________________________ >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