[BioC] question about edgeR and Poisson model: common dispersion is too low

Gordon K Smyth smyth at wehi.EDU.AU
Wed Apr 11 04:57:24 CEST 2012

Dear Galina,

Getting a small dispersion is good, not bad.  Your experiment is giving a 
bioligical CV of about 14% = sqrt(0.01859).  In our experience, that is 
typical or even on the high side for controlled experiments.  See comments 
in the discussion of the following paper:


Note that, while a dispersion 0f 0.019 may look small, it is still vastly 
different from assuming Poisson variability.

You give a quote from the first case study of the edgeR User's Guide. 
That data was from an observational study on human tumours, obviously 
highly variable data.

Best wishes

------------ original message --------------
[BioC] question about edgeR and Poisson model: common dispersion is too low
Glazko, Galina V GVGlazko at uams.edu
Tue Apr 10 16:26:03 CEST 2012

Dear list, dear Dr. Smyth

I am analyzing RNA-Seq data for the first time and I am confused with the 
value of common dispersion. I have only two replicates between conditions 
(A1,A2: normal; B1,B2: oxidative stress) and the value of common 
dispersion is

> d2<-estimateCommonDisp(d2)
> d2$common.dispersion
[1] 0.01859112
As far as I understand it is too low, is it?

In the User Guide  it is written that
'a common dispersion estimate of 0.2 means that there is a lot more 
variability in the data that can be accounted for by the Poisson model'

My estimate is much lower. Does it mean that I have to use Poisson model 
instead of NB? If I have to use Poisson model, is it possible to do with 
edgeR package?

I would appreciate your advices!
Best regards

> sessionInfo()
R version 2.15.0 (2012-03-30)
Platform: i386-pc-mingw32/i386 (32-bit)

[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United 
States.1252    LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                           LC_TIME=English_United 

attached base packages:
[1] splines   stats     graphics  grDevices utils     datasets  methods 

other attached packages:
[1] edgeR_2.6.0         limma_3.12.0        BiocInstaller_1.4.3

loaded via a namespace (and not attached):
[1] tools_2.15.0

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

More information about the Bioconductor mailing list