Ping Li posted on Saturday, September 29, 2012 - 9:16 pm
Hi Dear Drs. Muthen,
I have a binary independent variable for my MSEM model. After I run the model, M-plus report the error below:
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.420D-17. PROBLEM INVOLVING PARAMETER 14.
Parameter 14 is a binary independent variable. I also read other threads about this error. It seems a common error for binary independent variable. Should I pass it?
This message can be caused by including the mean and variances of a binary independent variable in the model. This is because the mean and variance of a binary variable are not orthogonal. You can test this by removing the variable and seeing if the message disappears. In regression, the model is estimated conditioned on the observe exogenous variables. So if you don't have a reason, they should appear only on the right-hand side of ON statements.
Ping Li posted on Monday, October 01, 2012 - 2:41 pm