[BioC] edgeR: regarding Poisson distribution and goodness of

[Yunshun Chen]

Hi,

1) If you remove the calcNormFactors() function, there will be no
normalization, which is not recommended in most cases.

2) To make a qq-plot for your Poisson data, the simplest way is as follows:

design <- model.matrix(~group)
fit <- glmFit(y, design=design, dispersion=0)
gof(fit, plot=TRUE)

3) To make the same graph with a NB distribution using tagwise dispersions,
you can do the followings:

y <- estimateTagwiseDisp(y)
fit2 <- glmFit(y, design=design)
gof(fit2, plot=TRUE)

Hope that helps.

Regards,
Yunshun Chen


------------------------------

Message: 9
Date: Tue, 18 Feb 2014 13:06:21 -0800 (PST)
From: "J [guest]" <guest at bioconductor.org>
To: bioconductor at r-project.org, jmillo4686 at gmail.com
Subject: [BioC] edgeR: regarding Poisson distribution and goodness of
	fit	graph
Message-ID: <20140218210621.8613C1437E3 at mamba.fhcrc.org>


Hello R and edgeR users/developers,

I had a question regarding the use of edgeR and graphing results. I'm trying
to do some comparisons between including and excluding different features in
edgeR. One variation I'm trying is the following:

x <- read.delim("fileofcounts.txt",row.names="Symbol")
 group <- factor(c(1,1,2,2))
 y <- DGEList(counts=x,group=group)
 et <- exactTest(y, dispersion = 0)

I believe this setup assumes the dispersion in my data is Poisson, and
calculates the gene-wise exactTest as so.  I've also removed the
calcNormFactors() function in this situation. So am I correct in suggesting
the only normalization that would be occurring in this case is with respect
to the library size? 

And my other question is how would I be able to make a qq-plot for this
procedure (goodness of fit statistics as the y-axis and Chi-square quantiles
as the x-axis)? The current gof() function appears to only be capable of
using GLM data as an input. Does anyone know how to do the same with Poisson
data that did not use the GLM functions? Furthermore, if I were to have
included the tag-wise dispersion it would also be good to know how to make
the same graph with a negative binomial distribution. So if anyone knows how
to make that graph I would be interested too. 

Thanks

 -- output of sessionInfo(): 

R version 3.0.2 (2013-09-25)
Platform: x86_64-apple-darwin10.8.0 (64-bit)

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
[1] LSD_2.5            ellipse_0.3-8      schoolmath_0.4    
[4] colorRamps_2.3     RColorBrewer_1.0-5 gtools_3.2.1      
[7] MASS_7.3-29        edgeR_3.4.2        limma_3.18.12     

loaded via a namespace (and not attached):
[1] tools_3.0.2

--
Sent via the guest posting facility at bioconductor.org.



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