Can someone recommend an approach for a gamm function in R that includes an autocorrelation (like the gamm(...,correlation=corAR1()) function in mgcv) but that is also recommended for handling binary response data? The gamm() help file has an explicit warning about using it for binary data.
"gamm performs poorly with binary data, since it uses PQL. It is better to use gam with s(...,bs="re") terms, or gamm4."
Best as I can tell, gamm4 currently doesn't have an implementation for autocorrelation.
Hopefully I'm just missing something obvious.
Thanks!