I have conducted an RNA-seq experiment with one factor in the design model, and the factor has several levels. The way I understand it from the vignette, I can use a global model with a likelihood ratio test to determine which genes are significantly regulated in any of the groups, compared to the rest. The code I use for this test is the following:
dds <-DESeqDataSetFromMatrix(countData = counts, colData =meta, design = ~ group)
deseq <- DESeq(dds, test="LRT", reduced = ~1)
res <- results(deseq)
My question is, does the order of the factor levels matter for the significance test? I don't really understand what the "full" and "reduced" models mean and whether the order of the factor levels can influence this.
Regarding the fold changes, I understand that the order of the levels determine which fold changes are presented by default, To my understanding, it makes more sense to use contrasts for determining fold changes between specific groups, but please correct me if I am wrong.
Any help would be appreciated!
