I would like to implement the two-part model of Olsen and Schafer (2001), who estimate coefficients in a model where the dependent variable is semicontinuous (lots of zeroes) and individuals are observed repeatedly over time. I guess my question is, is there a snatdard or easy way to do this, since the random intercepts estimated in the first part affect the covariance matrix of the effects in the second part (at least thats how I read it).
The first part is a logistic model with random effects. The second part numerically approximates a marginal likelihood w/ Gauss-Hermite quadrature, then maximizes it with Newton-Raphson.
I ended up going with PROC NLMIXED. A very helpful resource was the white paper by McTernan and Blozis, "User-specified likelihood expressions using NLMIXED and the GENERAL statement." Their first example was very close to what i was looking for. For future reference here is a link:
http://wuss.org/Proceedings13/67_Paper.pdf
This sounds like either fitting zero inflated models using PROC GENMOD or a mixture model using PROC FMM. Have you looked at the documentation for these procedures?
Steve Denham
I ended up going with PROC NLMIXED. A very helpful resource was the white paper by McTernan and Blozis, "User-specified likelihood expressions using NLMIXED and the GENERAL statement." Their first example was very close to what i was looking for. For future reference here is a link:
http://wuss.org/Proceedings13/67_Paper.pdf
I don't know if you can mark your own response as "Correct" in the new forums, but you certainly should, if able. Good finding.
Steve Denham
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.