If duplicates really are regularly spaced, then you should get the same consensus correlation whether you re-order or not. The fact that you get different results shows that something is wrong. The first thing is to check is that the duplicates are spaced as you think. Try unwrapdups(RG$genes$Name, ndups, spacing) This should give you a matrix with all columns the same. Does it? The second thing to check is that the assumptions listed below by Jan Oosting are satisfied. Best wishes Gordon On Wed, January 25, 2006 3:19 am, Hua Weng wrote: > Dear Oosting and Dr. Smyth: > > After I tried my regular spacing microarray data, I found that the common > coefficient after order with spacing =1 is different from the coefficient > with original spacing without order. And the difference is big, the first > one with order is 0.5618, the second one without order is 0.1711. Also the > B-statistics based on first one is only half of the value of the > B-statistics of the second one. Could you be kind enough to tell me why this > happens and is there some other good way can solve this irregular spacing > problem? > > Thank you very much, > > Hua Weng > > _____ > > From: Oosting, J. (PATH) [mailto:J.Oosting at lumc.nl] > Sent: Monday, January 23, 2006 3:50 PM > To: Hua Weng; Gordon Smyth; BioC Mailing List > Subject: RE: [BioC] limma: irregular spacing between within-array replicates > > > After normalization, which could rely on the chip layout, you can order your > data by gene, and assume a spacing of 1 > i.e. > MA<-MA[order(RG$genes[,"Name"]),] > corfit <- duplicateCorrelation(MA, ndups=4, spacing=1) > > this assumes that: > - there is a Name column in the gens slot that contains the same value for > each replicate spot > - each gene is replicated exactly 4 times > - if multiple probes are used for a gene they have different names