[BioC] edgeR: deDGE Fisher exact p-values

[Daniel Brewer]

Hi Mark,

Yes you are right I was using an old version of the vignette ... sorry
about that.  As for the design decision I have two suggestions:
1) A new function that simplifies exactTestNB (or modify exactTestNB
itself) so that you only have to provide an EBList-class object and the
groups (if more than two) and it returns a vector of results from the
exact test
2) If there is two groups provide an $exact element from deDGE, but if
not don't.

Thanks

Dan

Mark Robinson wrote:
> Hi Daniel.
> 
> Thanks for the note.  Indeed, there was a change in the package there as
> to where the exact p-values get calculated.  I have mixed feeling on
> this, but the decision to move this to topTags() was based on the fact
> that we now support multiple groups yet the exact test is only realistic
> for 2-group comparisons.  In topTags(), you must specify the pairwise
> difference of interest.  I would welcome your feedback on this design
> decision as like I said, I have mixed feelings of how this should be
> done.  The alternative is to precompute all the pairwise exact tests in
> deDGE() and maintain the $exact element.
> 
> As to the vignette example, I suspect you are using an older vignette. 
> The vignette for 1.2.x illustrates the same example with updated code.
> 
> Hope that helps.
> Mark
> 
> 
> On 12-Aug-09, at 8:57 PM, Daniel Brewer wrote:
> 
>> It use to be that when you run deDGE from the edgeR library that it
>> calculated the Fisher exact p-values and put it in a slot "exact".  This
>> no longer seems to be the case, with the p-values only being calculated
>> when topTags() is called.  This means that the plot example in the
>> vignette no longer works i.e.
>>
>>> adj.p <- p.adjust(ms$exact, "fdr")
>>> k <- (adj.p < 0.05)
>>> plotMA(ms, col = c("black", "blue")[k + 1])
>>
>> Have I missed something?  How do you get round this?
>>
>> Thanks
>>
>> Dan
>>
-- 
**************************************************************
Daniel Brewer, Ph.D.

Institute of Cancer Research
Molecular Carcinogenesis
Email: daniel.brewer at icr.ac.uk
**************************************************************

The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP.

This e-mail message is confidential and for use by the a...{{dropped:2}}