The classic way to test for better fit with an additional variable is to use the anova() function. The model must have the suspect variable listed last into your model. The anova() function will give you the correct sequential decomposition of your model effects and their conditional (F or t) tests. Check a regression text for the details. (You should have done this already.)
I have never heard of comparing residuals using the t-test. It makes no sense because the residuals have mean zero under either model. The AIC is also valid, but my reading between your lines would indicate the anova test would be better. JFL -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: Friday, September 14, 2007 9:49 AM To: r-help@r-project.org Subject: [R] Comparing regression models Dear list, I am interested in comparing two linear regression models to see if including one extra variable improves the model significantly. I have read that one possibility is doing an F test on the goodness-of-fit values for both models, and another option that is comparing the residuals of both models using a paired test. I also know about the anova() function that compares results for two models but am not sure what it actually does compare. Can you give me any suggestions? Does the same hold if the models were logistic instead of linear? I have read that the Akaike´s AIC is also a valid option. Thanks in advance for your comments David ______________________________________________ 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. ______________________________________________ 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.
