On Apr 23, 2005, at 4:44 AM, Gordon Smyth wrote: > >> Date: Fri, 22 Apr 2005 16:12:43 -0600 >> From: Jacob Michaelson <jjmichael@comcast.net> >> Subject: [BioC] covariance structures for limma >> To: bioconductor@stat.math.ethz.ch >> >> Hi All, >> >> With a *quick* search in the mailing list and through the limma PDF I >> couldn't find any information on how to provide for a covariance >> structure in limma (this is for a repeated measures experiment). I >> know lme has "weights" and "correlation" arguments which should do the >> job... > > There isn't any function called 'lme' in limma. Do you perhaps mean > lmFit()? Actually I was just referring to lme() in the nlme package... > >> can anyone enlighten me on limma's way of specifying covariance >> structures? > > limma supports several different covariance structures: see sections > on "technical replication", "within-array array replicate spots" and > "separate channel analysis" in the User's Guide. You probably want the > section on "technical replication". Only quite simple repeated > measures designs are supported. Thanks for the tip. I'll check it out...As an aside, are there any plans to support more complex repeated measures designs in the future? I know Bioconductor traditionally has covered more medicine-centric microarray experiments (where repeated meaures analysis doesn't come up too often), but being at an agricultural university with a new Affy core lab, I anticipate many more microbial and plant arrays with repeated measures designs. Fleshing out the repeated measures analysis capabilities in any BioC pacakge would be *very* helpful for current and future research as more and more ag and microbial researchers jump on the microarray bandwagon. > > Gordon > >> Thanks in advance, >> >> Jake Michaelson >