I have conducted some analyses using DESeq2 and obtained the output. However, instead of interpreting the effect sizes as log2FoldChanges, we would prefer to obtain a model-based mean difference.
I am trying to figure out how to calculate a "mean difference" from the log2FC and baseMean, and could use a second opinion: Does it make sense to use 2^log2FC * baseMean (maintaining the sign)? Since my data has a lot of variance (microbiome data), I'd rather use the variance shrunk estimates from the model. This is because when I take the mean for each group, I sometimes can get mean differences that are in the opposite direction of the log2FC.
Any thoughts would be appreciated. Let me know if more information would be helpful.