Hello all, I am looking for a quantitative measure of replicate quality to know if one should be tossed or not. Right now, its a gestalt from pairs and mva plots after running rma on datasets across all replicates and covariates. Would like your experience on reasonable method for quantifying "goodness of replication". Idea one: simply to use variance measure displayed in mva.pairs plots (This is function of my statistically ignorant empirical assessment- most replicates seem to fall <0.4 while between covariates is usually >.45). Alternate method might be use the slope of the fitted line on qq plots (or y intercept) of the M values? Does that measure sound at all reasonable? Is there any real difference between this and variance measurement above? Towards that end, I have been working with data in S-plus and making qq plots there on M values, is there easy way (ie function) within R to go from exprs values to qq plots of M values (or for that matter- to easily go from exprset object to MVA and QQ plots between covariates?) Tnaks in advance for your advice, John