

Mixture Model with covariates  warni... 

Message/Author 


I am trying to estimate a mixture model with covariates. An 'LCA only' analysis, without covariates, ran fine. Entering covariates leads to a problem. Here is the model: [...] Usevariables are x1x7 u1 u2 u3; categorical are x1x7 u1; Weight = dovwt2; Classes = c (7); Analysis: Type = mixture ; starts = 250 50; stiterations = 20; miterations = 2000; algorithm = integration; Model: %OVERALL% c#1c#6 on u1 u2 u3; [...] I receive the following warning: THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NONPOSITIVE DEFINITE FIRSTORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.931D12. PROBLEM INVOLVING PARAMETER 181. Further info: n = 8,605 The parameter identified in the warning is a threshold, which is estimated to be identical to its neighbour (i.e. the model estimates that there's an empty cell in one indicator category, in one class). May sparseness of information be the problem? The problem does not appear to be due to a particular covariate; it occurs as soon as I enter more than two covariates, whatever they are. I'd really appreciate any help if someone has an idea. 


Take u1 off of the CATEGORICAL list. This list is for dependent variables only. If you still have problems, send the output and your license number to support@statmodel.com. 


Thank you for your quick response, Linda. Unfortunately, taking u1 off the CATEGORICAL list did not solve the problem. Also unfortunately, my support contract has run out, and I am not currently in a position to buy a new one. I will seek advice elsewhere. Thanks again. 

Back to top 

