In theory, wouldn't they both show the influence that a variable exerts over an outcome?
To what extent those two methodologies are complimentary?
No they show different things. The independent variables in regression are measurements from nature. In FA they're computed theoretical constructions from the original IVs. The variables are not directly comparable to my understanding. Best, David
Hello Pedro,
Multiple linear regression and confirmatory factor analysis (CFA) both fall under the umbrella of the general linear model. Multiple linear regression tests the effect of multiple independent variables on one dependent variable that is measured continuously. Confirmatory factor analysis is a theory driven method used to see if items load onto domains or dimensions that may have previously been obtained from methods such as principal component analysis (PCA), or exploratory factor analysis (EFA).
A very simplified and applied example follows:
The WHOQOL-BREF is a quality of life (QOL) questionnaire that contains 26 items; 24 of the items load onto 4 domains, and the other 2 are a general subjective QOL item and a general health item. You could use a multiple regression to ascertain how the 4 domains predict either the QOL or health item, or (even better) perhaps a canonical correlation analysis to determine the effect of the 4 domains on both simultaneously.
Hypothetically speaking, someone using the WHOQOL-BREF in some previously unexplored sample could use PCA or EFA as an exploratory method to see if the item loading is identical to that found in the WHOQOL-BREF. If the loadings differ, one could collect another sample and use CFA to compare the divergent model to the original one. It should also be noted that the original domains of the WHOQOL-BREF were found using structural equation modeling, of which CFA is a subcategory.
Again, this is a very simplified answer tailored to your specific question. For more detailed discussions please see the references below:
Article The General Linear Model as Structural Equation Modeling
Applied Multivariate Research by Meyers, Gamst & Guarino.
All the Best,
Keston
Hello Pedro,
I think by multivariate linear regression you can only build a model. But you need CFA to confirm your hypothesis. Also, the factor loadings are different.
No this metods are diferents. The multivariate linear regression is a dependence technic and the CFA is a dependence and interdependence technic.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics