I'm writing a proposal, and am supposed to describe the statistics that I am likely to use; I do not yet have the data. I am investigating the use of two combined practices in education and for the quantitative aspect of the mixed methods approach, I have found two validated scales of 4-5 factors each, and not a combined scale, for each of the two sets of practices. There is likely correlation between at least some of the variables of both scales. As I develop a survey that uses factors from both scales, I anticipate that, similar to @Hope Derby's question from 2019 on here, I need to
1. Run and analyze the descriptive statistics
2. Run
--a multiple hierarchical regressions, for each of the outcome variables with control measures and predictor variables?
or
--a standard multiple regression with all predictor variables? but I'm considered about the possibility of multicollinearity
3. Report on the variance in R2 between questions from the two scales
Does that seem to fit what? Off to see, again, if any other studies have done this...
Hello Stephani Nikki Myers. Here are some resources I think you should review.
MANOVA
Article MANOVA: A Procedure Whose Time Has Passed?
Article Multivariate Analysis Versus Multiple Univariate Analyses
Both articles emphasize the importance of being clear about whether you have one multivariate question or multiple univariate questions. If you do have one multivariate question, the Huang article suggests considering SEM rather than MANOVA.
MULTICOLLINEARITY
- https://davegiles.blogspot.com/2011/09/micronumerosity.html
- https://www.statalist.org/forums/forum/general-stata-discussion/general/1652216-dealing-with-highly-collinear-independent-variables -- especially Clyde Schechter's posts
- https://www.econlib.org/archives/2005/09/multicollineari.html
- https://statisticalhorizons.com/multicollinearity/
HIERARCHICAL REGRESSION
You also mentioned hierarchical regression. I take that to mean that you will control which variables enter the model and the order in which they enter, and that those decisions will not depend on whether variables are statistically significant or not. But I think that some folks use it to describe what I would call some kind of stepwise selection of variables, where only statistically significant predictors enter or stay in the model. If you mean the latter, then here are a couple more resources I recommend.
I hope this helps.
If by hierarchical regression you mean controlling the order in which the variables enter the equation, this will have the same issues with multi-collinearity as standard multiple regression. I personally recommend hierarchical regression because of its increased interpretability.
In terms of having multiple dependent variables, you might consider Structural Equation Modeling rather than manova.
I think you first need to decide which multivariate measure provides predictors and which provides criteria or mediators, although it's possible the two sets are mixed in said provisions in any event if you follow @David L. Morgan's advice on using SEM, you need to come up with candidate models.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Materials Chemistry
- Polymer Chemistry