I am new to measurement invariance. I am using MPlus 7 to do a measurement invariance analysis between boys and girls. The data is categorical and I am using difftest, and theta parametrization. For my first step, I ran a configural model(which also has correlated error between two items on F2), the second step I constrained the loadings. I get a significant chi-square result. But when I examine the output, the loadings between boys and girls look very similar to me. The only difference is that for girls, the relationship (structural) for FI & F2 is not significant and for boys it is. I am confused as to how to interpret the data. Any direction would be appreciated.
I am also reading the Byrne, Shavelson, Muthen (1989) article about partial measurement invariance which appears to suggest that upon metric invariance, we still investigate structural invariance (relationships between the factors). Yes?
See the left column of the website. You will find the course videos there.
Testing for measurement invariance for categorical outcomes is not the same as what is described in the Byrne et al. article. See the current user's guide on the website under measurement invariance where the models for categorical outcomes are described.