[BioC] error message in edgeR

[Gordon K Smyth]

Dear Sridhara,

This is not an error message, it's a just an informative message that your 
data looks Poisson (dispersion=0) rather than negative binomial 
(dispersion>0).

Please note the help page for estimateSmooth(), which tells you that we do 
not recommend using this function.

I am currently recommending that you set prior.n = 30 / residual.df, where 
residual.df is the number of libraries subtract the number of groups. 
This has been made the default value now in the devel version of edgeR.

Best wishes
Gordon

> Date: Fri, 27 May 2011 21:37:10 -0400
> From: Sridhara Gupta Kunjeti <sridhara at udel.edu>
> To: bioconductor at r-project.org, bioconductor-request at r-project.org
> Subject: [BioC] error message in edgeR
>
> Hi,
> I got an below error message while running the edgeR.
>
>> d10 <- estimateTagwiseDisp(d10, prior.n = 10, grid.length = 500)
> Using grid search to estimate tagwise dispersion.
>> prior.n10 <- estimateSmoothing(d10)
> Warning message:
> In .odls(scores.g = scores, info.g = exp.inf) :
>  Estimate of overdispersion of likelihood scores is strictly zero.
> Returning min val close to zero (1e-08).
>
> I was not sure, what does this mean? Can anyone help me to troubleshoot
> this?
> I would great appreciate it.
>
> Thanks,
> Sridhara
>
> -- 
> Sridhara G Kunjeti
> PhD Candidate
> University of Delaware
> Department of Plant and Soil Science
> email- sridhara at udel.edu
> Ph: 832-566-0011

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}