Hello,
I'm working on creating latent profiles that will identify groups of survey respondents based on the age at which they report having experienced various life events. The data is survey weighted. I'm using Stata/MP 19.5
I've fit models with 1-5 classes and would like to evaluate their fit. However, when I request the usual statistics I use to evaluate latent class/profile analysis (estat gof, estat lcgof, estat ic) I get an error saying the command is not supported after svy estimation.
The only evaluation criteria I can obtain that does not produce this error is entropy, which to my knowledge is not a great choice for evaluating models relative to one another.
Are there better methods for producing goodness of fit statistics? If not, should I select a model based on non-statistical criteria (e.g., fit with theory/prior research)?
Thank you!
I'm working on creating latent profiles that will identify groups of survey respondents based on the age at which they report having experienced various life events. The data is survey weighted. I'm using Stata/MP 19.5
Code:
svy, subpop(included): gsem (ageevent1 ageevent2 ageevent3 ageevent4 ageevent5 <-, poisson), lclass(c 2) lcinvariant(none) covstructure (e._OEn, unstructured)
The only evaluation criteria I can obtain that does not produce this error is entropy, which to my knowledge is not a great choice for evaluating models relative to one another.
Are there better methods for producing goodness of fit statistics? If not, should I select a model based on non-statistical criteria (e.g., fit with theory/prior research)?
Thank you!