[BioC] multtest bug fixed

Katherine Pollard kpollard at soe.ucsc.edu
Thu Mar 31 00:42:15 CEST 2005


Hello All,

I've just found and fixed a bug in the boot.resample function of the
multtest package. This function is called by the function MTP. So, if you
have been using the MTP function with multtest version 1.5.x, you may have
incorrect p-values (rawp and adjp). Other functions, such as mt.minP,
mt.maxT and mt.rawp2adjp do not use this function and are OK.

The problem was that when standardizing the test statistics null
distribution, we used min where we needed pmin. The consequence was that
each marginal null distribution was standardized by 1/variance of the gene
with the largest variance (not its own variance). Hence, the marginal
test statistics null distributions were too narrow and more rejections
were made than should have been (ie: pvalues were smaller than they
should have been).

I have checked this correction into the cvs archive today and the fixed
package should be available under developmental packages soon. Version 1.6
(for the upcoming BioC release) will be corrected also.

Apologies for any inconvenience.

Best,
Katie



More information about the Bioconductor mailing list