An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20070817/ 4c0bbe33/attachment.pl

James, The underlying issues is that the interaction model requires all 4 cells to compute the interaction term. Essentially the interaction term is defined as a contrast between the four cell means. Since one does not exist, the interaction term is not estimable. Max -----Original Message----- From: bioconductor-bounces@stat.math.ethz.ch [mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of James Anderson Sent: Friday, August 17, 2007 6:12 PM To: bioconductor Subject: [BioC] question regarding unbalanced two way anova with interactionterm Hi, I am having trouble understanding the following: Suppose I have one gene with 26 samples. The first 20 samples are in batch A, the remaining 6 samples are in batch B. In the first 20 samples, there are 12 samples in control group and 8 samples in treated group, all batch B samples are in treated group. When I use a linear model (no interaction between batch and treatment), I can get some values out of it. However, when I use model with interaction, the variance are all 0, there must be something wrong with the model. I understand this is not a good experimental design, however, I don't understand why this kind of design will yield such kinds of results. Could any one explain to me why? Thanks, James --------------------------------- [[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 ---------------------------------------------------------------------- LEGAL NOTICE\ Unless expressly stated otherwise, this messag...{{dropped}}