LCA by groups: No observations error message

[Michelle Reiter]

Hi everyone!

I have done a LCA with 7 variables and 5 classes.

gsem (a1107_e a1113_b a1113_e a1113_f a1113_g_new a1113_h a1114_a <-), ologit lclass(C 5)

As my survey data consists of several countries, I would now like to see whether the coefficients vary over these different countries. However, I keep receiving an error message stating that there are no observations.

 gsem (a1107_e a1113_b a1113_e a1113_f a1113_g_new a1113_h a1114_a <-), ologit lclass(C 5) group(acountry) ginvariant(coef)
no observations
r(2000);

Could you explain to me what I did wrong, and how I could solve this problem? If I use the same code for gender instead of country everything works smoothly and I receive the latent classes per gender; no error message appears.
I have no missing values in my country-variable.

Thank you very much for every piece of advice!

[Clyde Schechter]

Perhaps country is a string variable?

[Michelle Reiter]

Thank you, Clyde
However, my country variable is numeric and not string.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(acountry asex a1113_h)
11 2  1
11 2  2
11 1  4
11 1  3
11 2  2
11 1  2
11 2  1
11 2  2

[Clyde Schechter]

Well, here's another guess. Perhaps for some level(s) of variable acountry, the estimation sample is empty due to missing values on other variables. Remember that if any of the model variables has a missing value, that entire observation is excluded from the estimation sample. So scattered missing values in the data can easily deplete an estimation sample. Try the following:

egen mcount = rowmiss(a1107_e a1113_b a1113_e a1113_f a1113_g_new a1113_h a1114_a)
by acountry, sort: egen estimation_sample_size = total(mcount == 0)
levelsof acountry if estimation_sample_size == 0

to get a list of countries that have no usable observations.