Drawbacks of Custom DE Methods

[DarioS]

In Chapter 6 Marker detection, redux there is recommendation against limma-style analysis. I wonder if that applies to today's times. One of the reasons is computational efficiency compared to Cohen's d. I often see data sets with fifty patients and if the gene abundances were averaged per gene and per cell group, perhaps limma would suit.

[alanocallaghan]

I think for a rigorous limma-style analysis it wouldn't be appropriate to average per cell group