Latent profile model including ordinal variables

[Charlene Rapsey]

I want to run a latent profile model including ordinal variables with 3 levels. I was sure that I had done this before but when I attempt to do this, I get an error message:

quietly gsem ( beaten hit parviol famment separate no_care <-, logit) ( Pnon Pcont Ppen <-,ologit), lclass(C 3)
option lclass() not allowed;
option lclass() is not allowed with models specified with continuous latent variables

Is anyone able to spot my error?

[Charlene Rapsey]

Is it the case that a capital letter tells Stata that I am using a continuous variable?

[William Lisowski]

It is never the case that capitalization tells Stata your variable is continuous. With sem and gsem, a captialized name is the default means of distinguishing latent variables from other variables.

To my reading, the error message tells you that you cannot have both a latent class variable (C) and continuous latent variables (Pnon, Pcont, and Ppen) in the same model. As I read your model, the variables Pnon, Pcont, and Ppen are considered latent, because their names are capitalized.

[Weiwen Ng]

To elaborate on William's reply, the correct syntax is

quietly gsem (beaten hit parviol famment separate no_care <-, logit) ( Pnon Pcont Ppen <-,ologit), lclass(C 3) nocapslatent latent(C)

-nocapslatent- turns off Stata's default setting, which is that any variable whose name starts in capital letters is a latent variable. -latent(C)- tells Stata that the only latent variable in your command is C. The -latent- option is required if you specify -nocapslatent-.

(NB: latent variables are generally continuous unless otherwise specified, but a latent class is a categorical variable. Stata can't currently mix continuous and categorical variables. Hence the error.)