dm484 posted on Sunday, December 13, 2009 - 6:33 am
Hi, I am running a path analysis with categorical variables, and the model looks like this (Y1 and Y2 are both binary variables):
... CATEGORICAL ARE Y1 Y2; ANALYSIS: ESTIMATOR=ML; MODEL: Y1 ON X1 X2 X3 Y2; Y2 ON X1 X2 X4 X5;
I met two problems:
1. The results of the two models were the same as the cases when I estimated the two equations separately. That is, if I don't use path analysis and estimate Y1 ON X1 X2 X3 Y2 and Y2 ON X1 X2 X4 X5 separately, the coefficients for the covariates are the same as those in the path analysis. I assume coefficients in path analysis or simultaneous equations should usually differ from those in separate equations, but don't know why this occurred.
2. There is a warning: 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 -0.952D-16. PROBLEM INVOLVING PARAMETER 5.
I don't know whether the problem of #1 is related to the warning in #2. Could you please give me some help? Thanks!
1. If I remember correctly from graduate school, with ML the parameter estimates are the same whether you run the univariate regressions separately or together. It is the standard errors that are different.
2. Number 2 is not related to number 1. Please send the full output and your license number for an answer to that.