Shuai Wang posted on Friday, March 12, 2010 - 11:56 am
I am running a path analysis with both continuous and binary variables. There are total 18 observable variables there, with five independent and 13 dependent (4 continous and 9 binary). There is no latent variable. All those 18 variables form a complicated path diagram. Paths start from those five independent variables, go through 17 dependent variables, and ultimately go to the same dependent variable. Number of observations are about 21,000, and I use WLSMV and theta parameterization.
After running the input program, I get the following message:
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL. PROBLEM INVOLVING PARAMETER 49.
I checked parameter 49. It is a beta between two binary variables. But I cannot delete that path without any evidence. Someone suggests trying log(x+1) transformation for all the continuous variables before estimation. I tried, but got the same error message. Is there anyone who can tell me how to remedy this? Your help is really appreciated.