Bert Gunter wrote: > > You can't. R^2 has no (consistent, sensible) meaning for nonlinear models. > If you don't understand why not, get local help or do some reading. >
There is no doubt that you are right, but Anna's question only shows that she is sandwiched between reviewers/supervisors who work in an applied field requesting an "objective" method; and the statistics community, who tells her that R^2 is nonsense and AIC is half-nonsense, as Frank Harrell points out frequently. I have similar cases to settle once a week, and they come up on this list with a similar frequency. When I try to argue with collections of r-list references like those compiled by David Winsemius, these are pushed aside as "cannot be cited". Trying to argue with some standard textbook on regression helps neither, because "we cannot read a whole book to find out why everyone requesting r^2 comparisons is wrong". So if somebody knows of some short reference in lucid applied-scientist language summarizing the facts, please post it here. I am looking for something on the level of Altman's papers in BMJ. Or some r-core Declaration of r-^2-Dependence. Dieter -- View this message in context: http://r.789695.n4.nabble.com/R-for-non-linear-model-tp3382397p3386649.html Sent from the R help mailing list archive at Nabble.com. ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.