[BioC] EdgeR - problems running estimateCRDisp

[Steve Lianoglou]

Hi,

I'm going to cut out most of your email to highlight the relevant part:

On Fri, Oct 22, 2010 at 1:44 AM,  <josquin.tibbits at dpi.vic.gov.au> wrote:
> I am running the new EdgeR package 1.8 and have updated R to 1.12.0 and
> updated all the packages I have installed.
>
> I have an RNAseq experiment with 24 samples and am wanting to run a glm
> analysis using the Cox-Reid common dispersion (and tagwise) paramaters.
> ...
> DGEList.object <- estimateCRDisp(DGEList.object, design)
> names(DGEList.object)
>
> This step is giving  the following error:::
>
> Loading required package: MASS
> Error: NA/NaN/Inf in foreign function call (arg 1)
> In addition: Warning messages:
> 1: glm.fit: algorithm did not converge
> 2: step size truncated due to divergence
>> names(DGEList.object)
> [1] "samples" "counts"
>
> Clearly the function is not working. I have tried two different computers
> and a run it in 64 bit but always with the same result.

Perhaps this is too simple a suggestion on my part, but do you have
any "such" values in your DGEList.object$counts matrix? And by "such"
I mean NA, NaN, or Inf values for some counts?

What does range(DGEList.object$counts) give you?

There's another description from the R-help archives of why this might
also happen, but I'm not sure how helpful that is to you at the
moment:

http://article.gmane.org/gmane.comp.lang.r.general/54844

-steve

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact