I am very new to doing LCA. I'm using Stata 15, in case that matters.

I am trying to do a latent class analysis using eleven variables, each with four levels of responses -- very important, fairly important, not very important, not important at all. There is no real theoretical guidance as to how many classes there will be, but my hunch is 3-4. I have run the below code with 2, 3, 4, 5, and 6 classes.

Code:
    gsem (var1 var2  var3 var4 var5 var6 var7 var8 var9 var10 var11 <- ) ///
[pweight=weight], ologit lclass(c 3) nonrtolerance
The code doesn't run without nonrtolerance. From my reading, if I was using binary instead of ordinal variables, I would next identify the class-variable combinations that had extremely high/extremely low probabilities and add, for example
Code:
(1.var1<- _cons@15)
There are some classes for which no respondents say "not important at all", which to me seems analogous to no respondents being 0 or 1 for a binary variable. But it's not obvious to me how to adapt the above code. Is +/- 15 still the magic number? Should I be referring to cut points instead of the constant? Or something else?