This video shows how to use Lavaan to conduct a multi-group CFA. It focuses on how to test loading (metric/weak) invariance and intercept (scaler/strong) invariance.
Thanks for your video! It was very useful! I have a question. At 2:10 , why did you choose the CFI coefficient on the left instead of the one on the right or the Robust CFI? Since you used maximum likelihood estimation with robust standard errors (MLR), I thought you would use one of those two. I'd glad if you could clarify that for me
This is a good idea for a video! I'm not sure how soon I'll be able to record one but here is some guidance whilst I get it together: (1) if you use fixed-factor method of identification then one group's mean will be fixed to zero and the estimate of the other group's mean will be the difference between the two groups. You can then fix that value to zero and use a chi-square difference test to see if the difference is statistically significant. (2) you can use Todd Little's effects coding method to generate latent means that are interpretable in the metric of measurement and create an equality constraint for the chi-square difference test. The two approaches are the same but the effects coding method gives you latent means to report.
This is very helpful, thank you!
Thanks for your video! It was very useful! I have a question. At 2:10 , why did you choose the CFI coefficient on the left instead of the one on the right or the Robust CFI? Since you used maximum likelihood estimation with robust standard errors (MLR), I thought you would use one of those two. I'd glad if you could clarify that for me
Can you please show us how to compare the two latent means after establishing the scalar invariance?
This is a good idea for a video! I'm not sure how soon I'll be able to record one but here is some guidance whilst I get it together: (1) if you use fixed-factor method of identification then one group's mean will be fixed to zero and the estimate of the other group's mean will be the difference between the two groups. You can then fix that value to zero and use a chi-square difference test to see if the difference is statistically significant. (2) you can use Todd Little's effects coding method to generate latent means that are interpretable in the metric of measurement and create an equality constraint for the chi-square difference test. The two approaches are the same but the effects coding method gives you latent means to report.
Dear sir why we do CFA MGA?
Can u share any references and any paper for this analysis?