Dear Mike, edgeR uses the term 'offset' in the same way that has been standard for generalized linear models since Nelder and Wedderburn's original paper in 1972. It agrees with the meaning in the glm() function and in thousands of books and articles on glms. It corresponds to your first formula for a glm with a log-link. One should be very careful about using established technical terms in a way that conflicts with the established definition. Best wishes Gordon ------------------ original message Wed Feb 20 10:11:45 CET 2013 Michael Love love at molgen.mpg.de [BioC] using offsets from EDASeq and cqn for edgeR But I think it would be desirable for packages on both sides of the hand-off to provide an explicit formula for what is meant by "offset", which could be one of: log(mu) = X * beta + offset log2(mu) = X * beta + offset log(mu) + offset = X * beta log2(mu) + offset = X * beta best, Mike ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}

Mike has raised a couple of important issues wrt. using the output from cqn as offsets in a GLM, for example by using edgeR. We are looking into this, and I'll make a follow up post when we have looked at this in detail. Kasper On Wed, Feb 20, 2013 at 6:05 PM, Gordon K Smyth <smyth at="" wehi.edu.au=""> wrote: > Dear Mike, > > edgeR uses the term 'offset' in the same way that has been standard for > generalized linear models since Nelder and Wedderburn's original paper in > 1972. It agrees with the meaning in the glm() function and in thousands of > books and articles on glms. It corresponds to your first formula for a glm > with a log-link. > > One should be very careful about using established technical terms in a way > that conflicts with the established definition. > > Best wishes > Gordon > > ------------------ original message > Wed Feb 20 10:11:45 CET 2013 > Michael Love love at molgen.mpg.de > [BioC] using offsets from EDASeq and cqn for edgeR > > But I think it would be desirable for packages on both sides of the > hand-off to provide an explicit formula for what is meant by "offset", > which could be one of: > > log(mu) = X * beta + offset > log2(mu) = X * beta + offset > log(mu) + offset = X * beta > log2(mu) + offset = X * beta > > best, > Mike > > ______________________________________________________________________ > The information in this email is confidential and intend...{{dropped:4}} > > _______________________________________________ > 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