I am fitting an LCA model. The model fit indices indicate that the model fits the data well. However, in my model results for one latent class, I encounter an estimate of -15.0000 for two thresholds and a corresponding two tailed p value of 999.
I understand that MPLUS somehow fixes this parameter when it estimates the LCA model and that this is not a problem as it just means the particular item Has a conditional response probability of 0 or 1.
My questions are:
1) How does MPLUS decide when to fix parameters in the LCA model estimation process?
As a follow up question to the above: it seems that the parameters fixed coincide with an item that was hypothesized to be predictive of latent class membership. Does this have anything to do with the choice of parameters being fixed by MPLUS or was this a coincidence?
2) What would be the appropriate way of reporting the output of this fixing of parameters for publication? (especially with regard to the parameter estimates and p-values)