when we use the modification indices to increase the model fit shall we make the relationship between different variable latent or should do the relationship among the variable latent?
Dear Logeswary Sooriyakumaran
Although the model is required to achieve satisfactory model fit, however, it should not be fully based on data driven decisions as using the modification indices. I would suggest you a set of action that you may take:
1. Take a look at the model fit indices of the other researches in your field and compare your model fit indices that your model has. If you find that they almost on the same level then I think it should be good. However, if not, proceed to point 2.
2. At the this stage, you should think of re-specification of your structural model to check if another structural model would result in a better fit, however, you should always be aware that the re-specification should also be logical, and if possible supported by a theory, a pervious study.
3. Use the fit indices to allocate the possible actions to take, however, once again, make sure that the decision taken is not fully driven by data.
Dear Logeswary Sooriyakumaran
If you are asking modification indices to increase the model fit in the measurement model, then you can only covary error terms that are part of the same factor.
- Badges
- Science method