Message/Author 

JW posted on Friday, October 26, 2007  8:08 am



Hi, I am wanting to test the effect of a moderator variable (gender) on a simple path model with single indicators (see syntax). The output provides the different results for the estimates, but how do I test if they are significantly different from each other i.e., males vs. females. My model is justidentified. Also, is this a sufficient method to test for moderation effect? Thanks! TITLE: FULL MODEl DATA: FILE IS rawdatagender.dat; VARIABLE: NAMES ARE DSRS SSSMINUS CNCEQ CESBQ SEX; MISSING ARE ALL (999); GROUPING IS SEX (1=MALE 2=FEMALE); ANALYSIS: TYPE=MISSING; ESTIMATOR IS MLR; MODEL: SSSMINUS CNCEQ ON CESBQ; DSRS ON SSSMINUS CNCEQ CESBQ; SSSMINUS WITH CNCEQ MODEL INDIRECT: DSRS IND SSSMINUS CESBQ; DSRS IND CNCEQ CESBQ; OUTPUT: CINTERVAL SAMPSTAT STANDARDIZED RESIDUAL; 


Multiple group analysis can be used to test for moderation effects. You can use chisquare difference testing to do this. This is described in Chapter 13 at the end of the section on multiple group analysis in the context of measurement invariance but the principles are the same. 


In the context of an MGASEM model I want to simply determine if structural coefficients are invariant across groups. I used code similar to that used in this threadi.e., no Model label" section. However, the values reported under Model Results (for unstandardized results) are not identical between the two groups. Do structural paths need to specified differently? Thanks! 


The measurement parameters of factor loadings and intercepts/thresholds are held equal across groups as the default. Structural parameters are not held equal. Inputs for doing this are shown at the end of the multiple group section of the Topic 1 course handout on the website. 

Back to top 