Calculating variance for microarray data? I have microarray data in which there are three biological replicates for each of the conditions. I am interested in a numerical estimate of how well the replicates correspond with one another, so I thought I should look at variance. I suppose I could calculate it individually for each probe, but I thought there might be more information by looking at all probes at once (something to do with moderation).
I've been poking around in limma, as a start, but I'm not sure how to go about this.
 A: It would be better to apply a calculation of variance to quantile ranked data considering the exons of genes rather than the whole microarray. It is the exons that should by far have the greatest expression. Antisense introns should have a low level of expression in most cases. Then if you take that number and compare it to the expression of the antisense-introns you will be able to judge how accurate or noisy is your data. This also tells you how the data varies from the prescribed localities of genes. Sometimes a noisy gene means bad gene location prediction. After you have validated you microarray then you can proceed to calculate the interprobe variance of exons and introns of genes. Variance of intergenic and unexpressed sequences is very difficult to prove as significant.
The quickest way to map your probes to genes will be to blast the probe database against the set of exons and introns or other material of interest be sure to turn filtering off. This method also identifies alternative splicing genes and many other features. 
Citation me its my published idea: Analysis of Antisense Expression by Whole Genome Tiling Microarrays and siRNAs Suggests Mis-Annotation of Arabidopsis Orphan Protein-Coding Genes
 
