[BioC] edgeR for large sample size

[Gordon K Smyth]

Dear Richard,

There's actually no problem in edgeR with prior.n less than one.  The 
remark in the User's Guide about keeping prior.n not less than one was my 
attempt to dissuade people from choosing prior.n too small when analysing 
small data sets.  It should probably be removed.

Best wishes
Gordon


> Date: Mon,  2 Apr 2012 14:02:47 -0700 (PDT)
> From: "Richard Hu [guest]" <guest at bioconductor.org>
> To: bioconductor at r-project.org, xuguang.hu at seattlebiomed.org
> Subject: Re: [BioC] edgeR for large sample size
>
> Hi edgeR expert:

>   I got a problem in using edgeR for RNA Seq data analysis.  The manual 
> recommends the use of getPriorN to get an appropriate estimate for 
> prior.n used in the algorithm and also warn that prior.n should never be 
> less than 1. Based on the recommended estimation formula prior.n * df = 
> 20 ~ 30 where df = #of samples - #of conditions, it seems that the edgeR 
> package will never work for sample size > 15 for two conditions if 30 is 
> used in the formula, and > 11 if 20 is used in the formula.

>   Am I wrong somewhere in understanding the package? I want to make 
> thing sure before I use it because I got more than 100 samples per group 
> in my RNA-seq dataset and the estimate of prior.n << 1. Please be clear 
> that my case is of 100 SAMPLES (or LIBRARIES) NOT 100 GENES or TAGS

>   I would appreciate any comments and helps from experts of edgeR
>
>   Best
>   Richard Hu
>
>
> -- output of sessionInfo():
>
> R version 2.14.2 (2012-02-29)
> Platform: x86_64-pc-mingw32/x64 (64-bit)
>
> locale:
> [1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252
> [3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
> [5] LC_TIME=English_United States.1252
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> loaded via a namespace (and not attached):
> [1] tools_2.14.2
>
> --
> Sent via the guest posting facility at bioconductor.org.
>

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