Dear all, 1- In normalizeCtData for norm="geometric.mean" , Can we average over all arrays or should we obrain the geometric mean of all samples? 2- geo.mean.ref= apply (a40set at exprs,1 , mean, na.rm=TRUE) gives an error but margin =2 does not produces an error, To me this is uncomprehensible. normalizeCtData(a40set, norm="geometric.mean",geo.mean.ref= apply (a40set at exprs,1 , mean, na.rm=TRUE))

Dear Ali, > Dear all, > > 1- In normalizeCtData for norm="geometric.mean" , Can we average over > all arrays or should we obrain the geometric mean of all samples? I don't really understand what you mean here. Can you define 'arrays' and 'samples' here, and what you're trying to do? Just to explain a bit more, 'geometric.mean' calculates the average Ct value across all samples, i.e. columns in the exprs matrix, and then scales each sample according to the sample average. In each case, a reference sample is indicated, i.e. the sample whose average is set to 1 in the scaling. This sample therefore remains unchanged after the normalisation. For example, using sample 1 as the reference: > data(qPCRraw) # Scale according to sample 1, i.e. the average Ct value of all # other samples are taken relative to this sample > test <- normalizeCtData(qPCRraw, norm="geometric.mean", geo.mean.ref=1) Scaling Ct values Using geometric mean within each sample Scaling factors: 1.00 1.06 1.05 1.02 1.04 1.02 # The refence sample remains unchanged > all(exprs(test)[,1]==exprs(qPCRraw)[,1]) [1] TRUE # The other samples are scaled > all(exprs(test)[,2]==exprs(qPCRraw)[,2]) [1] FALSE If using a different sample as reference, the scaling factors are now changed accordingly: > test <- normalizeCtData(qPCRraw, norm="geometric.mean", geo.mean.ref=2) Scaling Ct values Using geometric mean within each sample Scaling factors: 0.940 1.000 0.992 0.963 0.981 0.956 > 2- geo.mean.ref= apply (a40set at exprs,1 , mean, na.rm=TRUE) gives an > error but margin =2 does not produces an error, To me this is > uncomprehensible. > You don't say what the error is, so I can really comment on whether it's incomprehensible or not. However, none of the two should really work, since the input to geo.mean.ref is a single numeric value between 1 and the number of samples in your qPCRset. On a side note, rather than saying a40set at exprs, it's generally better to use the official accessor functions, i.e. exprs(a40set) or getCt(a40set). HTH \Heidi > normalizeCtData(a40set, norm="geometric.mean",geo.mean.ref= apply > (a40set at exprs,1 , mean, na.rm=TRUE)) > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor >