Hey,
I have a question regarding the structural equation model (appendix 1). As you can see, the factor loading from ImageGP to Intention has a value above 1, while the one from ImageGB to Intention has a negative value. I've read that factor loadings can be greater than 1, and that factor loadings can be negative (factor score indetermination?). However, having both these "issues" at the same time seems a bit odd. Is there any way to deal with these issues?
Best regards and thanks in advance,
Till
Try the following:
. Move from oblimin to promax rotation
. Recheck Scale Items Entry: There might be reverse coded scale with negative statements, that should be made converted to positive statements first. E.g., Five-point Likert scales (5+1-data response). . Drop Out the Statements: Statement should be dropped out
Hello Till,
What software are you using? If it is AMOS, please note that the program does not issue standardized loadings by default; you have to specifically ask for them (https://www.ibm.com/support/pages/how-can-amos-display-standardized-estimates-drawing-area)
If these are standardized estimates, then several possibilities exist for why values > 1 might occur:
1. You have a Heywood case (the observed relationships among variables are not coherent: small N, or pairwise deletion of cases are two possible explanations);
2. Some software packages require explicit setting of one of the links between manifest and latent variables to be set to 1, but this wasn't done;
3. The model as hypothesized is wrong;
4. See also this explanation from Karl Jöreskog (http://www.statmodel.com/download/Joreskog.pdf).
Good luck with your work.
Chung Tin Fah Thanks for your answer! I will try that!
Cristian Ramos-Vera Thank you!
David Morse Yes, I use Amos. Thanks for your reply and the insights, I will definitely use the paper by Jöreskog!
- Badges
- Science topic