Dear All,
I am trying to combine dependent p-values in R. May you please help me with
this?
For independent pvalue combination, one of the popular way is fisher's
method which I found the R code here
(http://r.789695.n4.nabble.com/fisher-s-posthock-test-or-fisher-s-combination-test-td2195964.html#a2305025):
fisher.comb <- function (pvalues)
{
df=length(pvalues)
ch2=(-2*sum(log(pvalues)))
return pchisq(ch2, df=df, lower.tail=FALSE)
}
For combination of dependent p-value, I could not find any R code here but
there is a method from Brown
(http://en.wikipedia.org/wiki/Extensions_of_Fisher%27s_method#Brown.27s_method),
in which it is still combine p value in the same way, the only difference is
the Chi square distirbution's variance is different from the previous method
(using the covariance of the p values).
I am not sure how to migrate from the traditional fisher's method to
brown's
method in R code. Because there is an introduction of covariance of the p
values. And the normal R function for pchisq seems do not take parameters to
change the variance of the chi square distribution. I am not sure if I fully
grasped the statistics behind it to write correct function for combination
of the dependent p-values. I hope someone having experience with this
problem can help me.
Thanks a lot in advance!
Cheers,
Shelly
--
View this message in context:
http://r.789695.n4.nabble.com/combined-dependent-pvalue-tp4647958.html
Sent from the R help mailing list archive at Nabble.com.
