I am trying to run a latent class model using the gsem command in Stata 16.1
I have 7 attributes of common beans of which five(5) are ordinal at 3 or 4 levels (cooktime1 taste2 climate2 yield2 price2) one is binary (swell2) and one is nominal (colour2) three types of colors of beans

my assumption is that a respondent's decision to choose a card in a discrete choice experiment is shaped by the number of people in the household (a18a), whether they use charcoal to cook beans or not (b4a_energy2), wealth index of the household ( wealth_index) kilograms of beans normally cooked (b6d) and whether household buys beans at least once a week or others (b6e_weekly)

The D-efficiency (using dcreate Stata command) estimated 84 choice scenarios which were blocked into 4 blocks. Each block had 7 choice cards with 2 options each and an opt-out option.

1. I have not read on this forum or anywhere how blocking is dealt with in such a case and seek assistance or reference to an old post if it exists.

2. The model has difficulty converging and would appreciate help in that regard

I am using the command below

gsem ( cooktime1 taste2 swell2 colour2 climate2 yield2 price2 <- ) (A <-a18a b4a_energy2 wealth_index b6c b6d b6e_weekly fcs_accept hdds_high wealth_index b6e_weekly b6b_ownprod), family(ordinal) link(logit) lclass(A 4) ginvariant(coef) lcinvariant(none)nonrtolerance startvalues(randompr, draws(40) seed(20) difficult) emopts(iterate(30)) intmethod(mvaghermite) technique(nr) from(b, skip)

3. I used the AIC BIC criteria to come to the 4 expected classes, but I can hardly distinguish the class characteristics. Maybe my classes are too many, is there another criteria to determine an appropriate number of classes in Stata.

I have attached the dataset

Thank you

NB: This is my first post on the run and had some difficulties write this post. Please understand if it's not well written. I can provide further clarification. And tips on writing better posts are very much appreciated.