[BioC] Bioconductor Digest, Vol 47, Issue 29

[Jenny Drnevich]

Hi Vinoy,

Please ask general questions like this on the Bioconductor list. It is very 
confusing at first, but the best place to start to get information on what 
the various functions are doing is to look at their help pages. For 
example, when you used 'write.fit', you were already making some decisions 
on how the p-values were calculated. The help file for 'write'fit' says 
"The p-value adjustment is done separately for each column. This is 
equivalent to decideTests with method="separate"." It all really depends on 
how you want to filter your data (... I hate the word 'filter'). Do you 
have more than one comparison (intercepts don't count)? If not, you don't 
have to worry about ClassifyTestsF, and you want method="separate" in 
decideTests. As far as what cut-off value to use for adjusted p-values, it 
all depends on what level of false-positives you are willing to tolerate. 
With the fdr/BH method, a common cut-off is 0.05, but I've seen 0.1 or 0.2 
used, which means that 10% or 20% of the genes in the significant list are 
expected to be false positives. Try to find a local statistician who would 
be willing to at least meet with you once to explain some of these issues - 
it's really much easier to do in person!

Jenny


At 08:27 AM 1/31/2007, Vinoy Kumar Ramachandran wrote:
>Hi,
>
>I have a basic question, i hope you can help me.
>
>I used limma for my time course arrays (common reference). I have written 
>the output data in excel file using write.fit with adjust (BH).
>
>What criteria shuld i take to filter the genes.......
>
>1) if  I use decideTests, different number of significant genes come up 
>with different flavours of decideTests.
>
>2) ClassifyTestsF
>
>3) adjusted p. value or the q value.
>
>how to decide on the cut off value of decideTests or classifytestsF or in 
>q value. what factors shd i consider to decide them.
>
>I have gone through the limma user forum and read the posts, but stil i am 
>not getting a clear idea about using them. I am now learning statistics.
>
>Thanking in advance
>
>with regards,
>vinoy

Jenny Drnevich, Ph.D.

Functional Genomics Bioinformatics Specialist
W.M. Keck Center for Comparative and Functional Genomics
Roy J. Carver Biotechnology Center
University of Illinois, Urbana-Champaign

330 ERML
1201 W. Gregory Dr.
Urbana, IL 61801
USA

ph: 217-244-7355
fax: 217-265-5066
e-mail: drnevich at uiuc.edu