Is it possible in MPLUS to get in a path analyses besides lagrange multipliers (on adding a relationship), also a Wald test (on deleting relationship)? It would be nice to obtain that, but I can't find how to do this...
M.G. Keijer posted on Monday, November 09, 2015 - 1:28 am
Dear Linda Muthen,
I am running a multiple group model for males and females. The are 3 free pads and 50 constrained paths for both sexes. One of the free paths is significant in the female model at a level below 0.01 and in the male model the same path is not significant (above 0.1). When I run a Wald test on this i.e. a1=b1 (df=1), I get a non significant result.
The same happens for other models I run or father and mothers, etc.
It hard for me to get my head around the idea that 1 coefficient might be zero, the other one is not zero, but do they differ? No!
Could you please help me out? Thank you in advance.
The Wald test is not testing against zero. It is testing the coefficients against each other. Even though one is different from zero and the other is not, they still may not be different from each other.