after EFA , 5 latent variables are extracted....after getting factor scores for these 5 latent variables... can I use one of those variables as dependent variable in multiple linear regression....
Asked
Active
Viewed 396 times
0
-
https://stats.stackexchange.com/q/51814/3277 and.... – ttnphns Jul 15 '18 at 10:55
-
2Possible duplicate of [Factor analysis and regression](https://stats.stackexchange.com/questions/29824/factor-analysis-and-regression) – ttnphns Jul 15 '18 at 10:56
1 Answers
1
It is said that this approach typically introduces bias. One reason I can think of is that factor scores do not have a perfect correlation with the latent variable. Besides the factor scores can be extracted in many different ways yielding different values for your regression model. So approaches that keep it all in the model are preferable. Structural equation modeling is one way to go. Multilevel modeling (with Bayesian software required to vary loadings by items) is another way, if you can set it up. If you still intend to extract the factor score prior to doing the analysis, then see this recent paper. It includes formula at the end for bias correction.
Hypothesis testing using factor score regression: A comparison of four methods.
Heteroskedastic Jim
- 4,567
- 1
- 10
- 32