[BioC] globaltest Error when using subset data

[J.J.Goeman at lumc.nl]

 Dear Xia,

This is a recently discovered bug that occurs in the permutation version
of the test when there are empty genesets (which have empty intersection
with the set of gene names). As globaltest automatically selects the
permutation version of the test when there are few samples, the bug only
occurs when you run the subset analysis.

You can try either using the asymptotic p-values:

gt.kegg.day8 <- globaltest(as.matrix(rma.data[ , 9:20]),
treatment[9:20], kegg, method = "asymptotic") 

This probably the best choice as you also had asymptotic p-values in the
full analysis.

Alternatively, you can remove the empty genesets:

empty <- sapply(kegg, function(path) {
	length(intersect(path, geneNames(rma.data))) > 0 
})
kegg <- kegg[!empty]

I hope this helps you out. A bug fix will appear in the devel version in
a few days.

Best,

Jelle

 

> -----Original Message-----
> From: Xia Han [mailto:xia.han at bms.com] 
> Sent: 15 February 2007 17:06
> To: bioconductor at stat.math.ethz.ch
> Subject: [BioC] globaltest Error when using subset data
> 
> Hi:
> 
> I am trying to run globaltest with an affy data set containing 30
> mouse430a2 chips. These are the codes I am using:
>  > library(globaltest)
>  > library(mouse430a2)
>  > kegg <- as.list(mouse430a2PATH2PROBE) # treatment is the 
> factor indicating different treatments (containing 3 time 
> points which didn't seperate here).
>  > treatment
>  [1] Control Control Control Control CD137   CD137   UC10    UC10    
> Control Control Control Control CD137   CD137   CD137 
> [16] CD137   UC10    UC10    UC10    UC10    Control Control Control 
> Control CD137   CD137   UC10    UC10    UC10    UC10  
> Levels: Control CD137 UC10
>  > gt.kegg.all <- globaltest(as.matrix(rma.data), treatment, kegg)
> 
> Everythig was fine when I ran all the dataset together as 
> above. However when I try to use subset data and rerun the 
> globaltest, I got following
> error:
>  > gt.kegg.day8 <- globaltest(as.matrix(rma.data[ , 9:20]), 
> treatment[9:20], kegg) Error in cbind(deparse.level, ...) : 
> number of rows of matrices must match (see arg 2)
> 
> I've wasted a lot of time trying to fix this error with no 
> luck. Please help!
> 
> Thanks,
> 
> Xia
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> 
>