|
 |
Likelihood function specification for... |
 
|
Message/Author |
|
Jimi Oke posted on Tuesday, August 01, 2017 - 9:10 am
|
|
|
Hello Drs Muthen and staff, Please could you provide or point me to the specification/formulation of the exploratory factor analysis model with both continuous and censored variables? I've checked a few of your papers but cannot seem to find any explicit documentation of the expectation vector and the log-likelihood function. In particular, I would just like to know exactly what is being computed in the current version of Mplus (8.0) for Monte Carlo EFA with censored variables specified along with the continuous ones. Thanks, Jimi |
|
|
See the article 29) Muthén, B. (1989). Tobit factor analysis. British Journal of Mathematical and Statistical Psychology, 42, 241-250. [Available as PDF] which you find at http://www.statmodel.com/bmuthen/full_paper_list.htm |
|
Jimi Oke posted on Tuesday, August 22, 2017 - 9:08 am
|
|
|
Thanks Dr. Muthen. Thanks for your kind response, and sorry for the delay in getting back. I had seen that article in the past, but I just wanted to make sure that this was still the current implementation. The reason was because, in the console output, I would see a reference to a Poisson model for the censored variables and wanted to verify. I would assume the [zero-inflated] Poisson model is used for the "count" variables, right? So, maybe in that instance there was a misspecification somewhere. I also wanted to ask if there is a limit to how many censored variables one may specify in an EFA. Our models seemed stable when we had up to 26. But when we tried to go further, the rest were truncated and treated as continuous. Best, Jimi |
|
|
Yes, Poisson refers to outcomes specified as counts. Censored is something else. There is no limit on the number of censored variables. If you see strange results, please send output to Support along with your license number. |
|
Back to top |
|
|