Hello!
I used Sawtooth to run a latent class analysis on a 5-task CBC with 4 attributes and a none option. The highest number of levels in an attribute was 7.
For academic purposes I would now like to replicate this analysis in R and SAS (if possible). I have PROC LCA from Penn State's Methdology center, but I'm not sure if or how that can be used with the design specifications instead of just the choices on each task (which are devoid of meaning without the specifications, since the design had 300 versions). I have also read about using PROC PHREG to run conditional MNL, but I don't know if or how to use that to detect latent classes.
Likewise, I have looked at various packages in R, but I am encountering similar conceptual roadblocks.
My basic questions are:
1) is it possible to incorporate information about the design of each choice task into estimating latent classes in SAS and/or R? What PROCs and packages can be used? What format does the data need to be in? Do the variables identifying each level need to be dummy coded? If so, how does the software know which levels belong in each attribute?
2) how would I transform the beta coefficients into zero-centered diffs?
Any advice, code, existing resources, etc are greatly appreciated!
p.s. As I said, this is for academic purposes - literally. I am writing up these methods for a presentation for a biostatistics class for my PhD program in epidemiology!