Is there any literature that supports deleting of more than 20% of indicators to achieve model fit in SEM. I have managed to achieve NFI ,833 , RFI ,825, IFI ,915, TLI ,911, CFI ,915 , RMR ,050 , CMIN/Diff 1.862 after deleting 12 indicators out of 30 in one construct. Kindly advise
I would not recommend doing this. Please consult a statistician for details. Best wishes David Booth
@Munuhwa As you mentioned that you have 30 observed items under one single latent construct which is way too many (in my opinion) to load appropriately under one single construct and thus leading to misspecification of the model.
Hi, I suggest not to delete so many indicators . I think, you can try to use item parceling to achieve model fit.
You can take a look at the following papers:
Landis, R. S., Beal, D. J., & Tesluk, P. E. (2000). A comparison of approaches to forming composite measures in structural equation models. Organizational Research Methods, 3(2), 186-207.
Yang, C., Nay, S., & Hoyle, R. H. (2010). Three approaches to using lengthy ordinal scales in structural equation models: Parceling, latent scoring, and shortening scales. Applied Psychological Measurement, 34(2), 122-142.
Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F. (2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural equation modeling, 9(2), 151-173.
Rogers, W. M., & Schmitt, N. (2004). Parameter recovery and model fit using multidimensional composites: A comparison of four empirical parceling algorithms. Multivariate Behavioral Research, 39(3), 379-412.
@Imran Anwar thank you for observing that i guess the observed items are too many on one construct. @Wang Zheng thank you so much , im trying to figure out how i can do item parceling and see the outcome.
Wang Zheng , item parceling really helped, i managed to achieve model fit without deleting even a single item from the construct.
- Badges
- Science method