Dear Michael,

I am using DESeq2 to analyze differential expression between two conditions (RNAseq, 3 replicates each). When playing around with different fit types as parameters for the DESeq() function I found that the default (parametric) fit results in 29 genes with adjusted p-values below my threshold while the local fit gives me 153.

This lead me to the question which of the two I should use and if there is an objective way to make this decision. I obviously don't want to choose one just because it gives me more/better results.

The plots I get from the plotDispEsts() function can be found here:

local fit: http://i.imgur.com/grpWtnv.png

parametric fit: http://i.imgur.com/A79woce.png

It would be great if you could give me some suggestions as to how this decision should be made? If at all possible, I'd like to find a way to assess the fit in a more quantitative way than visual inspection.

Thanks in advance,

Christoph

Hi Michael,

although I understand the principal of this approach I am having troubles implementing the computation into R. Could you share some code snippets how to compute the median of the absolute residuals for the following example?

Thanks already in advance :)

The residuals are: