My problem is regarding a structural model with six latent (independent) variables. After running CFA and when I do the structural model, all hypothesis were rejected. But if I put 1 as regression weight for one of the latent variables' regression paths, some hypothesis are significant. I want to know that assigning regression weight for regression paths is correct or not?
Dear Chamari,
Adding constrains on regression paths or other estimated parameters should be always theory driven. You can not just put 1 because it makes other estimates significant.
Regards,
Marina
Your question is unclear. Are all these latent constructs (IV and DV) being measured by multiple items? Usually, one would restrict one of the factor loadings to 1 in the CFA model, for identification. If you are estimating the structural model along with the measurement model, you should continue to do that, but restricting one of the regression paths to 1 does not make sense.
Using 1 as regression weight for one of the latent variables' regression path should be supported by literature review or expert opinion. On another note, sometimes researchers feel frustrated because of negative results or insignificant results, however, we understand that this is also a contribution because now we have an opportunity to identify what is/are causing something new that was not usually expected.
- Badges
- Science topic