Simulate datasets with classified observations and associated categorical variables

Question

I need to simulate a bunch of datasets where the variables are categorical variables characterized by some dependency patterns that I decide (They are perfectly independent, perfectly dependent, some variables are conditionally independent give the others and so on). I also need to have as a result a "true" classification for each observation. I know that there exist the function in the package poLCA in R that does that, however, the variables in that case are independent, and I haven't found any way in order to specify the correlation/association between them. Does somebody know another way to do so?