Hello,
In my project, the criterion has two subscales (which are really important to me), can I have two subscales for the dependent variable while using hierarchial regression in SPSS?
Thanks in advance :)
Hello Sara,
Strictly speaking, least squares regression (whether simultaneous or hierarchical) only involves one dependent variable (DV) at a time. If you really want to include two DVs, perhaps you could think about using canonical correlation, which can evaluate optimal combinations of one set of variables to maximally correlate with another set of variables. Here's a good resource for some of the basics about CC:
Sherry, A., & Henson, R. K. (2005). Conducting and interpreting canonical correlation analysis in personality research: A user-friendly primer. Journal of Personality Assessment, 84, 37-48. doi:10.1207/s15327752jpa8401_09
The other choice is structural equation models, which allow you to propose and evaluate simple or highly complex sets of relationships among variables (both directly measured and inferred, latent variables/factors).
The best choice likely depends on the specific research question(s) you're trying to answer.
Good luck with your work!
Before doing anything else, I would examine the correlation between your two subscales. If it is quite high (e.g., .8 and above), it will be difficult to find statistical differences, even if you believe that there are substantive differences. Of course, you can "eyeball" these differences by running parallel regressions separately for each DV, but as David Morse says, you will need to run more advanced models to make a statistical comparison of the two models.
Personally, I would recommend Structural Equation Modeling rather than canonical regression In particular, SEM will allow you test the whether each of the coefficients linking the independent to the dependent variables are equal -- as well as the global test that all of them are equivalent.
FYI my preference for SEM is that it is much better known in the social science fields in which I work, but if people do routinely use canonical regression in your own field, that would be a recommendation for it.
Ali Hameed Yousef They are independent sub-scales, but I DO want to compare them with eachother.
Thanks a lot David Morse . That was really helpful. Since I have a mediator variable, I am thinking of Structural Equations, I hope I can handle all its complexities. :)
David L Morgan Well, I am also working in social science (psychology), I think I will choose Structural Modeling. Thanks a lot for your help.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS