[BioC] EdgeR exactTest - treatment names determine sign of fold change

[Spollen, William G.]

Yes, I saw that. But in the context of a loop where you make the 6 pairwise comparisons of 4 group members it becomes important to name your groups in either asending or descending order to keep the fold changes consistent.  I just wanted to alert anybody else in my situation to this possible source of trouble.

Thanks
Bill

________________________________________
From: Mark Robinson [mark.robinson at imls.uzh.ch]
Sent: Saturday, May 12, 2012 2:20 AM
To: Spollen, William G.
Cc: bioconductor at r-project.org
Subject: Re: [BioC] EdgeR exactTest - treatment names determine sign of fold change

Hi Bill,

You can control this with the 'pair' argument in your call to exactTest().  Have a look at the description in ?exactTest

Best,
Mark

----------
Prof. Dr. Mark Robinson
Bioinformatics
Institute of Molecular Life Sciences
University of Zurich
Winterthurerstrasse 190
8057 Zurich
Switzerland

v: +41 44 635 4848
f: +41 44 635 6898
e: mark.robinson at imls.uzh.ch
o: Y11-J-16
w: http://tiny.cc/mrobin




On 11.05.2012, at 22:12, Spollen, William G. wrote:

> Hi
>
> I am still new to EdgeR and am using it to do pairwise comparisons of RNA seq data.  I just realized that the sign of the fold change will be determined by how you label the samples in groups.  So if I have T1 and T2 as my two groups, it is labeled in that order in d$samples, and in the levels reported in d$samples$group.  If I use as names T80 and T60 d$samples will list them in that order but the levels in d$samples$group will sort them and that changed the sign of the fold change. I am using the same input file each time.  All else is the same, I just changed the labels I used for the samples.  The code for exactTest seems to bear this out, using the levels of d$samples$group to determine how to take the ratio.
>
> Thought I should report that as I was looping through a list of comparisons and wondered why  the sign changed.  I am using v2.6.2 EdgeR.
>
> Thanks
>
> Bill
>
>
>       [[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor