[BioC] (stupid) question about wilcoxon test and finding interesting genes

Naomi Altman naomi at stat.psu.edu
Tue Feb 15 21:56:54 CET 2005

When there is no differential expression (and if the genes were 
independent) then the p-values should be uniformly distributed.  So, if you 
test at level alpha and you have N genes, you SHOULD find alpha*N genes 
that have significant results (and all are false positives).

The FDR correction does 2 things simultaneously - it estimates the 
percentage of genes that differentially express (using departures of the 
p-values from the uniform distribution" - and then estimates the False 
discovery rate for any observed p-value.

I guess we have to ask what "necessary" and "multiple testing" 
mean.   There are 2 kinds of error - false detects and false 
non-detects.  We do not do this type of correction if we worry more about 
false non-detects.  If false detects are a bigger problem, then the FDR 
estimate allows us to estimate when we have an acceptable rate.  If you are 
really testing only a few genes on your arrays, I would not use FDR.  If 
you are really testing all the genes, then I think you have a "highly 
multiple" testing situation.

I don't really like the term "adjusted p-value" for FDR estimates.  They 
are not probabilities, they are estimated error rates.  But that issue was 
discussed a few weeks ago on this list.


>so the next step is to correct the p values... i thought correcting p 
>values is only necessary when i do multiple testing? sorry for my 
>question, but i am more used to do some programming and work with 
>databases then doing statistics...
>>>Bioconductor mailing list
>>>Bioconductor at stat.math.ethz.ch
>>Naomi S. Altman                                814-865-3791 (voice)
>>Associate Professor
>>Bioinformatics Consulting Center
>>Dept. of Statistics                              814-863-7114 (fax)
>>Penn State University                         814-865-1348 (Statistics)
>>University Park, PA 16802-2111

Naomi S. Altman                                814-865-3791 (voice)
Associate Professor
Bioinformatics Consulting Center
Dept. of Statistics                              814-863-7114 (fax)
Penn State University                         814-865-1348 (Statistics)
University Park, PA 16802-2111

More information about the Bioconductor mailing list