Assume there's two factors that affect the metabolite compounds content of my data: Factor A and Factor B, Factor Ahas three level and Factor B has two level, so there is 3*2 groups, and all in one experiment batch. I only want to get the compounds related to the Factor A, however Factor B seems to have dominate effect on the data(from unsupervised dendrogram and PCA).
I have tried two approaches to analysis differential compounds: First way is use ANOVA to analysis full model(content ~ Factor A + Factor B) and reduced model(content ~ Factor A and Factor B). Second way is use multiple t.test to analysis different group of Factor B, and after p-adjust, use the overlap of their differential sets. The third approach is to use permutation F test set p < 0.05 and use F cutoff to estimate FDR and make FDR< 0.05 to get the differential compound set.
I tried both ways and get the differential compound sets(p<0.01, FDR<0.05), however, on the dendrogram the same species samples still cluster together using first or third approach, and the second approach which use multiple t.test, since the sample size is two small there is significant result.
Is there any other better ways to do this?
And thanks in advance for your advice :)