Hi at all,
i have a quite general question on how to model a SEM and i hope someone can help me out.
So here is the situation:
In our institute, we obtained a dataset, trying to measure the innovation capacity of students. This has been done by a previous work, where 9 sub-scales have bin summed up to a single factor which results in the innovation capacity of students. We measured two times, at the beginning and at the end of the first academic year.
I know want to build a SEM, where the innovation capacity after the first academic year is basically the "dependent variable". As "independent variables", there are on the one hand observed variables (items in the questionnaire) which i put in to the SEM model using the "insert observed variable" rectangle. But i also have some other "constructs" (latent variables), which result through CFA of different items in the questionnaire. The problem is: how should i set up the model:
Method 1: Should i build factor scores for the innovation capacity (by building factor scores for each of the 9 sub scales and the load those 9 factor-score "variables" into one singe factor) and insert this into the SEM by using the "add observed variable" icon OR...
Method 2: Should i insert the single items that "build" the latent variables e.g. innovation capacity with the "add observed variable" icon and then draw single paths to the latent variable icon (circle)?
The reason i ask this is, when i use method 2, the gof measures are way worse than using method 1. I'm just wondering if method one is a valid way of doing this analysis.
Hope you can help me out, thank you very much in advance!
i have a quite general question on how to model a SEM and i hope someone can help me out.
So here is the situation:
In our institute, we obtained a dataset, trying to measure the innovation capacity of students. This has been done by a previous work, where 9 sub-scales have bin summed up to a single factor which results in the innovation capacity of students. We measured two times, at the beginning and at the end of the first academic year.
I know want to build a SEM, where the innovation capacity after the first academic year is basically the "dependent variable". As "independent variables", there are on the one hand observed variables (items in the questionnaire) which i put in to the SEM model using the "insert observed variable" rectangle. But i also have some other "constructs" (latent variables), which result through CFA of different items in the questionnaire. The problem is: how should i set up the model:
Method 1: Should i build factor scores for the innovation capacity (by building factor scores for each of the 9 sub scales and the load those 9 factor-score "variables" into one singe factor) and insert this into the SEM by using the "add observed variable" icon OR...
Method 2: Should i insert the single items that "build" the latent variables e.g. innovation capacity with the "add observed variable" icon and then draw single paths to the latent variable icon (circle)?
The reason i ask this is, when i use method 2, the gof measures are way worse than using method 1. I'm just wondering if method one is a valid way of doing this analysis.
Hope you can help me out, thank you very much in advance!
Comment