Hello,

During the analysis of my data, I found while estimating the over-dispersion parameter of the gene ratios (using bbEstDisp ; in the apeglm package) that some of gene almost reach the upper limit (pretty much regardless of the limit). Should such gene be discarded as this parameter is not accurately evaluated ? Should gene almost reaching the minimal limit also be discarded ? Should this be done post or pre evalutation of the s-value ?

Thank you

relevent material

The upper limit just means the counts are close to binomial-distributed. So don’t discard.

You’ll notice that you can barely tell the difference in the pdf of a beta binomial as the ‘theta’ parameter grows very large.