Hi,
It is highly appreciated to help me with the model (attached word file). The data sample size is 200. As the attached word file shows the result of Baseline Comparisons is not acceptable and is so low, even if the loadings of the items are not low. The field of study is related to work psychology and employee behaviour. Based on the literature review the connection between the variables are logical. Any idea for improving the model fit is so appreciated.
Thanks
Shaghayegh
Some results are acceptable, while others are not. CMIN/df and RMSEA are acceptable., while CFI and others are less than .90 or .95. I wonder what the variable which was in the rectangle. I would suggest to try to increase the sample.
So it looks like you have a serial mediation model with multiple higher-order latent variables. With a sample size of only 200, the model is likely too complicated to run, as also demonstrated by the fit indices. You could opt to instead run partial least squares SEM if the purpose is pretty much for prediction rather than theory testing or comparison (and if you want to keep latent variables). This will at least help to save you the headache of dealing with fit measures.
Dear Shaghayegh, Your results of the SEM are acceptable. However you need to increase CFI and TLI and also decrease RMSEA. In my opinion it is better to take a look on your factor loadings. In my opinion by removing those measurement items that are below 0.5 you can have better fit indices. in addition to this pay more attention to modifications that are recommended by the AMOS software.
Hohjin Im
Thanks for your helpful recommendation. I will try to rerun the model with partial least squares.
Saeid Nosrati
Thanks for your sharing the information. The problem with the second-order construct remained unchanged even after removing the items with low loading and doing modification indices.
The result is only satisfactory with observed variables.
- Badges
- Science topic
- Similar topics
- Filing