Testing for coefficient differences b...
Message/Author
 Thomas Zerback posted on Tuesday, November 05, 2013 - 9:21 am
Dear Drs Muthen,

I'm running two separate models using the same sample. The models only differ in one latent dependent variable (MK / PR, both have the same metric):

Model 1:

EM by EG01_01 EG01_03G;
EM on F1_ARGUD;
EM on F2_UMFRD;
MK by MK01_01G MK01_03G;
MK on F1_ARGUD;
MK on F2_UMFRD;
MK on EM;

Model 2:

EM by EG01_01 EG01_03G;
EM on F1_ARGUD;
EM on F2_UMFRD;
PR by PR02_01 PR02_02;
PR on F1_ARGUD;
PR on F2_UMFRD;
PR on EM;

I want to test if the differences between the path coefficients are significant, namely:

"MK on F1_ARGUD" compared to "PR on F1_ARGUD"
"MK on F2_UMFRD" compared to "PR on F2_UMFRD"
"MK on EM" compared to "PR on EM"

I already used the "model test" command, but I'm still uncertain if it makes sense in my case:

EM by EG01_01 EG01_03G;
EM on F1_ARGUD;
EM on F2_UMFRD;

MK by MK01_01G MK01_03G;
MK on F1_ARGUD (p1);
MK on F2_UMFRD (p2);
MK on EM (p3);

PR by PR02_01 PR02_02;
PR on F1_ARGUD (p1b);
PR on F2_UMFRD (p2b);
PR on EM (p3b);

model test:
0 = p1 - p1b;
0 = p2 - p2b;
0 = p3 - p3b;

 Bengt O. Muthen posted on Tuesday, November 05, 2013 - 4:03 pm
MK and PR need to have at least metric invariance, that is, the loadings need to be equal. You say the two factors have the same metric but I don't know what you mean by that. If you have different factor indicators it seems hard to argue invariance. But Model Test is right in principle.
Post:
This is a private posting area. Only registered users and moderators may post messages here.