Simon Wood's `gamm()` function in the mgcv package allows one
to fit smoothing terms through the gam() functionality and random
effects through the `lme()` functionality.

- The output from the gam portion of the model is stored in
`mod$gam`and the lme portion in`mod$lme`(assuming`mod`is the name of the`gamm()`model object. - Note that
`predict(mod$gam)`and`predict(mod$lme)`give the same output, namely the predictions of interest. - One use of this is that one can define one's own variance function.

Last modified 12/28/07.

Chris Paciorek 2012-01-21