[BioC] Why edgeR ouput some p values larger than 1

[Gordon K Smyth]

Dear Yuan,

exactTest() does not produce p-values larger than one as far as I am 
aware.  For example I just ran:

> library(edgeR)
> example(exactTest)
> summary(de$table$p.value)
    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
  0.1295  0.4351  0.6770  0.6201  0.8535  1.0000
> max(de$table$p.value)
[1] 1

If you think there are p-values>1, please give an example.  You can't 
judge from a histogram, because the right bar of the plot may appear to 
extend past 1 even when the values themselves don't do that.

Best wishes
Gordon

> Date: Thu, 5 Jan 2012 14:24:21 -0800
> From: Yuan Tian <ytianidyll at ucla.edu>
> To: <bioconductor at stat.math.ethz.ch>
> Subject: [BioC] Why edgeR ouput some p values larger than 1
>
> Hello,
>
> I was using edgeR to do differential expression on a RNA-seq dataset, 
> but when I check the p.value (not adjusted) distribution of the output, 
> I found ~700 genes got a p value larger than 1... Just wondering if 
> anyone else got this problem before? And why I got such a weird a 
> result. Here's my codes,
>
> d <- DGEList(counts=countsTable, group=conds, lib.size = lib.sizes) #I used the raw read counts as instructed in the user guide
> d <- calcNormFactors(d, method="TMM")
> d <- estimateCommonDisp(d)
> et.commonDis <- exactTest(d)
>
> hist(et.commonDis$table$p.value, breaks=100, col="skyblue", 
> border="slateblue", main="pval distribution_edgeR")
>
> May I have your ideas? Thanks.
>
> Yuan

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