[BioC] Filter over group averages

Wolfgang Huber whuber at embl.de
Thu Mar 11 16:17:32 CET 2010


it is not difficult to write such a function, using "split" on your 
covariates, and then "lapply" and "rowMeans" on the resulting list of 
sets of arrays.

The reason such a function does not exist in genefilter is that this 
type of filtering disturbs the null distribution of subsequently used 
t-statistics or F-statistics that are commonly used in per-gene 
hypothesis testing, and thus typically leads to invalid p-values. See 
the man page of the "nsFilter" (and the reference cited therein, which 
will appear soon).

If you don't do hypothesis testing afterwards, i.e. if you use this 
filtering as a variable selection step e.g. for classification, 
clustering, then that is of course fine.


Michael Imbeault scripsit 08/03/10 18:53:
> Hello,
> Is there a way to easily filter genes on the average signal over groups 
> (according to phenoData?). Let's say that in phenoData of my eset object 
> I have 'Time' and 'Status' variables, which have 3 values each. I have 3 
> arrays for each of the (9) conditions.
> What I want is to keep genes with an average value of at least 100 in at 
> least one of the resulting 9 conditions. The genefilter package 
> seemingly only include functions to filter genes ignoring groups - is 
> there an easy way to do the same taking into account groups?
> Thanks,
> Michael
> _______________________________________________
> 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


Best wishes

Wolfgang Huber

More information about the Bioconductor mailing list