I have a score of 3 dependent variables in continuous form and 1 Independent variable in also continuous form, I want to see the effect of 1 independent variable on those 3 independent variables, I can do that in simple regression analysis by putting that in three combinations one by one. But I wanted to do it in a single execution. MANOVA gives such leverage but the problem is independent variable has to be non-metric or categorical. In my case, my independent variable is metric or continuous so should I still use MANOVA or is there any other option? or do I keep my metric independent variable in the covariate option given the MANOVA dialogue box in SPSS and run the test? If I am doing that then I am getting the parameter estimate similar to simple linear regression which I did making a combination one after the other. Am I going in the right direction?
Hello Pathak,
It's most important here to determine what your specific research question might be, rather than to try to apply some specific statistical method to your data set.
Can one execute a manova with one continuous IV and three continuous DVs? Yes, treating the IV as a covariate, in the language of manova.
But, you can get the same information by running a single OLS regression model, using the three variable set as the IVs, and the one continuous variable as the DV.
Why? Because manova generates an optimal linear combination of the DVs to maximize the explanatory power of the IV (whether continuous, categorical, or some combination when multiple iVs are used). The regression model would explicitly report that linear combination and be directly interpretable, with respect to both the coefficients and the amount of shared variance.
Good luck with your work.
Pathak Abhijit sent me essentially the same question via email. Here is the reply I sent.
Hello. I suggest that you read the classic article by Huberty & Morris (1989). Even though you have a metric explanatory variable rather than a categorical variable, the question they pose must be considered: Do you have one multivariate question or [three] univariate questions. If you decide you have one multivariate question, you can estimate your multivariate model using GLM. The code would look something like this: GLM Y1 Y2 Y3 WITH X /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /PRINT=PARAMETER /CRITERIA=ALPHA(.05) /DESIGN=X. The output will contain both multivariate and univariate tests, but again, you should choose which approach you want in advance, and ignore the other. https://www.researchgate.net/publication/232494414_Multivariate_Analysis_Versus_Multiple_Univariate_Analyses
HTH.
@Bruce Weaver Thank you sir once again, yes sir I will read it again. I have one multivariate question. I just wanted the multiple feedback from avid researchers, that's why I posted this question. I will go through the Andy Field book on SPSS to get more insight.
See also:
http://psychologicalstatistics.blogspot.com/2021/08/i-will-not-ever-never-run-manova.html
Article MANOVA: A Procedure Whose Time Has Passed?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression