[BioC] Filtering Post-Normalization With limma

[Matt Jones [guest]]

Hi,

Chapter 7 of the limma User's Guide introduces the concept of filtering after normalization. I'm not entirely sure how this step should be performed. Is this simply the removal of probes flagged as poor quality by the scanner/quality assessment criteria? If so, should all spots associated with this probe be removed across all replicates or should it simply be the spot in question specific to that array? The former approach would effectively remove the probe from the experiment but the second would simply mean that the size of the sample used to fit the linear model for that probe would be reduced by one.

Of course, maybe neither of the above approaches are correct. If this is so, could somebody provide some insight for this step. The user's guide suggests, "one way is to keep probes that are expressed above background on at least n arrays, where n is the smallest number of replicates assigned to any of the treatment combinations." What does this mean?

Thanks.

Matt

 -- output of sessionInfo(): 

No sessionInfo().

--
Sent via the guest posting facility at bioconductor.org.