[BioC] Using eBayes to find P values for individual contrasts

[Mark Cowley]

Hi Jason,
you normally generate one topTable per contrast of interest. so if you  
have 3 coefficients of interest, you would call topTable 3 times, each  
time choosing coef=1, then coef=2, then coef=3 in 3

I usually just write a for loop for each coef of interest & store the  
topTables in a list.

cheers,
Mark
On 18/11/2010, at 1:58 PM, Jason Shoemaker wrote:

> Dear Mark,
>
> Thank you for the warning! I was worried about asking something  
> silly. So if I may ask, how can I get topTable to display not just a  
> single adjusted p value for one contrast, but the adiusted P values  
> for all contrasts? I don't seem to see this option. Thus I have been  
> applying p.adjust to the raw P values to adjust the values for each  
> contrast of interest.
>
> Thank you!
> Jason
>
> On 11/17/2010 10:32 AM, Mark Cowley wrote:
>> Hi Jason,
>> I think you're in danger of reinventing the wheel.
>>
>> The adj.P.Val column in the topTable is the corrected p value.  
>> Don't forget about the coef topTable parameter to control which  
>> coefficient to look at. You can control what method to use via the  
>> adjust.method parameter.
>>
>> then take a look at the decideTests method to work out which genes  
>> are significant  for which contrasts.
>>
>> cheers,
>> mark
>>
>> On 16/11/2010, at 7:28 PM, Jason Shoemaker wrote:
>>
>>> Dear all,
>>>
>>> I have searched the archives but not found any advice on this  
>>> issue. I am using the LIMMA package to determine differentially  
>>> expressed genes. I have been using eBayes plus topTable to find  
>>> the most differentially expressed genes, but now I want to  
>>> determine the adjusted p values specific for each contrast. Should  
>>> I simply do as follows (following the example from http://matticklab.com/index.php?title=Single_channel_analysis_of_Agilent_microarray_data_with_Limma) 
>>> :
>>>
>>> contrast.matrix<- makeContrasts("Treatment1-Treatment2",  
>>> "Treatment1-Treatment3", "Treatment2-Treatment1", levels=design);
>>> fit2<- contrasts.fit(fit, contrast.matrix)
>>> fit2<- eBayes(fit2)
>>>
>>> P.values<-p.adjust(fit2$p.values,methods="fdr");
>>>
>>> in doing so- can I make fair comparisons between p values for each  
>>> contrast? Or more precisely, if a get a p value of 0.01 for  
>>> "Treatment1-Treatment2" and large value (P>0.1) for the remaining  
>>> 2 contrasts, is that gene significant only for "Treatment1- 
>>> Treatment2"?
>>> Thank you!
>>> Jason
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>>
>>
>> -----------------------------------------------------
>> Mark Cowley, PhD
>>
>> Peter Wills Bioinformatics Centre
>> Garvan Institute of Medical Research, Sydney, Australia
>> -----------------------------------------------------
>>
>>
>