[BioC] dispersion in edgeR

[Naomi Altman]

I have 2 questions about dispersion in edgeR.

1) The model implies that as sequencing depth increases, the power 
for testing differential expression comes to an asymptote.  This seems odd.

2) Usually when using a shrinkage estimator, values of the original 
estimates shrink monotonely towards the common estimate.  So, if one 
plots the moderated values against one another for 2 values of the 
shrinkage parameter, the plot should be monotone increasing.  The 
plot was too big to attach,  but what I did was:

d10=estimateTagwiseDisp(d,prior.n=10)
d30=estimateTagwiseDisp(d,prior.n=30)
plot(d10$tagwise.dispersion,d30$tagwise.dispersion)

I have not included my particular set of data, as I am pretty sure we 
see this for any set.

This plot seems to imply that 2 genes could have the same moderated 
dispersion values at prior.n=10 and very different values at 
prior.n=30.  This is not due to my

--Naomi


Naomi S. Altman                                814-865-3791 (voice)
Associate Professor
Dept. of Statistics                              814-863-7114 (fax)
Penn State University                         814-865-1348 (Statistics)
University Park, PA 16802-2111