Hi all, I'm trying to do a paired 3control vs. 2experimental chip analysis with SAM, and my class label is : sam.cl <- c(-3, -2, -1, 1, 2) but this throws an error : Error in adjust.for.mt(data, cl, var.equal = var.equal) : Negative values are only allowed for the paired analysis. Or, if a paired analysis should be done: There is something wrong with the class labels. So what should I do? Is it best to just drop one of the control chips so that I have an equal number of control and experimental chips??? (It was planned as a 3x3 comparison, but our core botched one of my experiemental chips). Thanks, Ken

Uh... how can it be a paired test if you don't have 2 measures for each unit? In my humble opinion I wouldn't use a permutation test with such a small sample size, as your simulated null distribution would be quite coarse. But witn a 2 vs 2, which actually is a n=2 sample size (being paired) I don't think you can do that much statistics. HIH, Stefano On Fri, Apr 22, 2005 at 03:23:25PM +0000, Ken Termiso wrote: <ken>Hi all, <ken> <ken>I'm trying to do a paired 3control vs. 2experimental chip analysis with <ken>SAM, and my class label is : <ken> <ken>sam.cl <- c(-3, -2, -1, 1, 2) <ken> <ken>but this throws an error : <ken> <ken>Error in adjust.for.mt(data, cl, var.equal = var.equal) : <ken> Negative values are only allowed for the paired analysis. Or, <ken>if a paired analysis should be done: There is something wrong with the <ken>class labels. <ken> <ken> <ken>So what should I do? Is it best to just drop one of the control chips so <ken>that I have an equal number of control and experimental chips??? (It was <ken>planned as a 3x3 comparison, but our core botched one of my experiemental <ken>chips). <ken> <ken>Thanks, <ken>Ken <ken> <ken>_______________________________________________ <ken>Bioconductor mailing list <ken>Bioconductor@stat.math.ethz.ch <ken>https://stat.ethz.ch/mailman/listinfo/bioconductor