Hello everyone, I need to make a design matrix, and I can use the approach outlined in 3.3.1 of the edgeR users guide, but I think that there may be a better way to conduct the analysis. I have 36 samples in a 3x2x2 incomplete factorial design. I have three genotypes (wild type, a single mutant, and a double mutant), two tissues (roots and shoots), and two treatments (no trt, trt). the mutants are m1 and m1/m2. One approach would be to group the treatment groups into a single variable Group, as is done in the edgeR guide, and then follow the example, extracting interaction terms to find the differences between each genotypes response. There should be a different way to model this problem that is more / as powerful as the previous method (right?). Because in the double mutant the gene expression is going to have a component that is dependent on each mutation and then the interaction of each mutation. So I am interested in trying to estimate the genes that are responsive to treatment and dependent on the introduction of mut2 into the mut1 background (for tissues independently) Does such a method exist, how can I do it / what should I read to understand how to do it? Thanks, -- Sam McInturf [[alternative HTML version deleted]]