Two-part mixed model for longitudinal data

[Ha DN Le]

Hi all,

I'm analysing healthcare costs of a longitudinal cohort study. My outcome var (cost) is continous and has excess zero. Just wondering whether there is a two-part model command for longitudinal data. I've used the twopm command for cross-sectional analyses but not aware of the command for longitudinal data.
Pls help if you have any hints. I don't know SAS so not interested to learn another software b/c have seen the code for two-part model in longitudinal data in SAS.

Many thanks,

Ha

[Clyde Schechter]

You can accomplish this with finite mixture modeling. -help fmm-. The zero inflation can be accommodated with a -pointmass- distribution in -fmm-. If, as I would expect for healthcare cost data, the remaining data is highly skew to the right, consider using a poisson model for the other component. If you open -help fmm- and click on the "View complete PDF manual entry" link near the top of the window, and then to go Remarks and Examples, you will see an example of this in Example 3.

Added: Never mind, that's not for longitudinal data.