JW posted on Thursday, November 01, 2007 - 6:13 pm
Hi, I'm running a just-identified path model and want to test for moderation effects using a categorical grping variable, with an MLM estimator.
So, i've read the MGA section in the users guide. I know I have to calculate the SB chi test, but I'm confused as to which values are the To and T1 as well as the cd. i can't find the cd in my output (the users guide says it provides this automatically?). The output also provides two chi sq values, one for the baseline model (which doesn't change across groups) and the another chi sq which does change between the equal form group and the restricted group. Which do i use?
Another question is that when I compare my equal form model to my more restricted model, do I restrict parameters in both the overall model and the group specific model? And then compare this chi sq to the chi sq in the equal form (unrestricted model)? I was informed by a colleague to restrict path X>Z and then look for changes in Chi and then look at paths Z>Y separately, and again look for changes in chi sq. Example syntax for the Z>Y path: ANALYSIS: ESTIMATOR IS MLM; MODEL: SSSMINUS CNCEQ ON CESBQ; DSRS ON SSSMINUS CNCEQ (1); DSRS ON CESBQ; MODEL INDIRECT: DSRS IND SSSMINUS CESBQ; DSRS IND CNCEQ CESBQ; MODEL MATDEPHX: DSRS ON SSSMINUS CNCEQ (1);
T0 is the chi-square value for the nested model. T1 is the chi-square value for the less restrictive model which will have a chi-square of zero and no scaling correction factor. The scaling correction factor in the nested model is found with the chi-square test statistic. You don't want the baseline model.
I think your colleague is saying to test one restriction at a time. You might also consider the Wald test using MODEL TEST.