[BioC] Fisher's method of combining p-values

[Adaikalavan Ramasamy]

The statistics from Fisher's method simply sums up the p-values on log 
scale across the studies and multiplied by minus two. Under the null 
hypothesis, this follows a chisquare distribution with degrees of 
freedom equivalent to twice the number of studies.

If you store the p-values from different studies as columns in 
'pvalues.mat' where the rows are genes. Then if I remember my 
arithmethics correctly you have

  stats    <- rowSums( -2*log( pvalues.mat ), na.rm=T )
  Npresent <- rowSums( !is.na(pvalues.mat) )
  pval     <- pchisq( stats, df=2*Npresent, lower=FALSE )

Please double check.

Regards, Adai




Laura wrote:
> Hi,
> Is there an implementation of Fisher's method of combining p-values
> available as a part of package ?
> 
> thanks,
> lau
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> 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