Relevant crossposting. Win98, S-PLUS 2000 (Sv3), library(mix). I've estimated a GLM for each multiply imputed data set that I've got. Then I've used mi.inference to combine estimates and standard errors for each coefficient. Now I have the coefficients and the standard errors in a matrix. How do I get them "back" into a GLM object ? I need this to be able to use predict on the final model. The GLM is a logistic regression.
I can assign the coefficients themselves back into the model using model$coefficients <- mi.model[,"coeffs"] but I can't force the standard errors in using the same syntax. I'd like to be able to treat the final model as a "real" GLM so I can use predict. The model contains a splined covariate, thus the coefficients themselves aren't that amenable to interpretation. I've looked at the S-NEWS archives, the online manuals, MASS and Basics of S-PLUS. Any ideas? Leafing through the source code for summary.glm and predict.glm it doesn't seem that difficult, since the dispersion parameter in my case is 1. But I'm still not sure how to do it. Any ideas? Yours, Jan Brogger, MD, PhD student, University of Bergen, Norway
