[BioC] edgeR and tagwise dispersion: overcorrection for multiple tests?

[Gordon K Smyth]

Dear Alessandro,

The phenomenon you observe is not generally true, and it is does not 
represent "overcorrection for multiple tests".  It is likely to be 
specific to your data set.

Please note that we do not recommend the use of common dispersion except 
for interpretation purposes.  We always recommend tagwise dispersion.

If you were to use common dispersion for differential expression when the 
true dispersions are heterogeneous, and I believe that they generally are, 
then you would under-estimate the dispersion for some genes, leading to 
overly small p-values, and over-estimate for others, leading to overly 
large p-values.  So it would be perfectly possible to get a few very small 
p-values before correction for multiple testing but get less significance 
after adjustment for a particular data set.

The bottom line is, use tagwise dispersion unless you have just one 
residual df.

Best wishes
Gordon


> Date: Wed, 04 Apr 2012 16:30:06 +0200
> From: "alessandro.guffanti at genomnia.com"
> 	<alessandro.guffanti at genomnia.com>
> To: bioconductor at r-project.org
> Subject: [BioC] edgeR and tagwise dispersion: overcorrection for
> 	multiple	tests?
>
> Dear all, I am new to this list so kindly forgive me if this question
> has been already answered -
> but in this case I would be grateful if you could point me to the right
> answer
>
> I am using edgeR for DEG analysis of SAGE tags and I noticed that, when
> using tagwise instead of common dispersion, the number of differentially
> expressed genes raises if I consider uncorrected P-values.
>
> However, apparently the number of genes called siginficant with
> BH-corrected drops significantly
>
> Is there a rationale behind this or it is just my impression ?
>
> Thanks in advance and kind regards,
>
> Alessandro G.
>
>
> -- 
>
> Alessandro Guffanti - Head, Bioinformatics, Genomnia srl
>  Via Nerviano, 31 - 20020 Lainate, Milano, Italy
>     Ph: +39-0293305.702 Fax: +39-0293305.777
>             http://www.genomnia.com
> "When you're curious, you find lots of interesting things to do."
> (Walt Disney)

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