mmlcr is an Splus/R library for mixed-mode latent
class regression (also known as mixed-mode mixture model
regression or mixed-mode mixture regression models) which can
handle both longitudinal and one-time responses. By mixed-mode, I
mean that the manifest variables can be of mixed types: some
longitudinal, some one-time, some normal, some censored-normal,
some categorical, some Poisson. The software is written
completely in Splus code and is built around the EM algorithm. It
is slow but very flexible.
Right now there are three forms of the library. One is a zip-ed library, version 1.1, for Splus 2000 (Windows machines). Place the mmlcr.zip file in the "library" folder of Splus, and unzip it. There are Windows help files to get you started. The other is a compressed tar file containing a "dump"-ed file of the source code, plus the Unix-style help files.You'll have to turn it into a library yourself.
For SPlus 6 on Windows, there is a new version, 1.2, of mmlcr. Place mmlcr.zip in the "library" folder of SPlus, and unzip it. The Unix version can be found at mmlcr.tar.gz. Move to the library directory and run the INSTALL script. (Note: I don't actually have access to SPlus 6 for Unix, so I haven't been able to test on this platform.)
For either of the Splus versions, if you used censored normal variables you'll have to fix in bug in survReg; see the file README.TXT in the distribution.
Finally, the most up-to-date version is for R, mmlcr_1.3.2.tar.gz .
This version has many improvements over the Splus versions, and is the only
one I'm currently working on.
Please send any comments, questions, and
suggestions to me, Steve Buyske.
Back to my main web page.