I am studying the relationship between business process orientation (BPO) and cross-functional integration (CFI). BPO is formed of five distinct dimensions. Having done factor analysis two factors were also extracted for CFI, meaning that I have now two correlated DVs. My initial thought was to run two separate multiple regressions, one between the five IVs and one CFI factor (DV) and another one between the five IVs and the second CFI factor (DV). The other alternative I have found from the literature is Canonical correlation test used when there are two or more dependent variables, but it has received some criticisms due to the complexity involved in interpreting the results.
FYI, BOP is a second order latent variable with five constructs reflective in nature, and CFI is also a second order latent variable with two factors extracted but formative in nature. Attached provides a better visual explanation of my final framework.
Experts advice would be much appreciated on the two above alternatives. Many thanks!
When there are more than one dependent variable which are correlated and both of them are dependent upon the same set of independent variables, then we can use the method of multivariate regression which allows multiple dependent variables provided assumptions of linear regression are satisfied. The method can be tested using statistical software packages like SPSS or others.
When there are more than one dependent variable which are correlated and both of them are dependent upon the same set of independent variables, then we can use the method of multivariate regression which allows multiple dependent variables provided assumptions of linear regression are satisfied. The method can be tested using statistical software packages like SPSS or others.
Structural equation modeling (SEM) in Amos SPSS software or Multivariate regression analysis is appropriate.
I like to handle all model parts at the same time (i.e factor and predictor models simultaneous). For this you should consider a structural equation model. In my opinion the easiest program to work with is Mplus (www.statmodel.com). It is designed for models involving the kind of path relationships among observed and latent variables shown in your model figure. Mplus has a simple interface language, powerful optimization engine for conducting analyses and a hefty set of examples in the user's guide (http://www.statmodel.com/download/usersguide/MplusUserGuideVer_8.pdf), and it is capable of handling also other outcome measurement scales beyond the continuous (e.g. various nominal and ordinal types as well as counts and survival times). There is a free demo available, but it is restricted to a limited set of variables that can be included in the model (http://www.statmodel.com/demo.shtml).
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running