

Factor alignment method possible? 

Message/Author 

Eric posted on Tuesday, February 18, 2014  7:17 pm



Dear all, I have a question about using the factor alignment method. I’m analyzing data from an intervention study. The design of the study is as follows: There are two groups, group 1 (n=332) received an intervention whereas group 2 (n=217) served as a control group without an intervention. The intervention should change subject’s beliefs about certain topics. All groups were administered several scales to measure the beliefs, scales where given at two time points, one measurement before and one measurement after the intervention. Altogether, I have four measurements. The hypothesis is that the factor means for group 1 differ for the two measurements, whereas for group 2 they should be the same. I first considered fitting a longitudinal model (with correlated errors), one separate model per group in order to compare the factor means. But the scales only show configural invariance thus rendering mean comparison not feasible. The models with invariance restrictions didn’t yield acceptable fit. But apparently the factor loadings are only marginally different, it is only one or two items per scale having different factor loadings at the two time points. My question is now if I could use the alignment method for the comparison of the two measurements? The webnote on this method mention only the comparison of many groups but not longitudinal measurement. Best regards, Eric 


Alignment is for groups that are independent, not different time points for the same people. You can try partial invariance for just some items. 

Eric posted on Wednesday, February 19, 2014  4:42 am



Bengt, thank you for clarification and your hint! 

Back to top 

