

Cannot find "results in probability s... 

Message/Author 

Reem Saeed posted on Wednesday, October 16, 2013  6:09 am



Hello, My sample size consists of 118 areas. These are considered IDs and I have included them in the analysis section as auxiliary. Although the best likelihood estimate is replicated and the model terminated normally. I still get this error: 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 0.967D16. PROBLEM INVOLVING PARAMETER 83. Can this message be ignored? I also cannot see the results in probability scale to be able to label the classes appropriately. Any help and advice is very much appreciated Thanks 


Please send the output and your license number to support@statmodel.com. 

Reem Saeed posted on Monday, October 21, 2013  5:12 am



Dear Linda, I have now come up with results from my LPA. Please note that my variables are proportions, therefore they range between 0 and 1. I have done a 3 class and a 2 class model. The 3 class model has a better fit according to the BIC: 2 class model BIC > 9965.662 3 class model BIC > 10613.238 However, for both models I get the following error message: 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 0.140D16. PROBLEM INVOLVING PARAMETER 16. My first question is: 1. Is it alright or normal to get negative AIC and BIC values? 2. Can this message be ignored? Please note that first I had a similar message involving parameter 22, I deleted it and reran the analysis. I still get the same message, but with a different parameter. 3. My variables are slightly skewed, could this be causing it? What are the implications of running a LPA with slightly skewed continous variables? and what would you suggest to solve this issue? Many thanks, Reem 


1. Yes. 2. Please send the output and your license number to support@statmodel.com. 3. This is not a problem for mixture models. 

Back to top 

