In a multiple regression, we have three dependent variables (A, B and C) and we have five independent variables (V, W, X, Y and Z).
At the time of checking the Collinearity statistics of A, B and C, the value of tolerance and VIF of five independent variables (V, W , X, Y and Z) are same for all dependent variables (A, B and C) .
As an example
tolerance and VIF of (V, W ,X,Y and Z) with A are (0.716, 0.964, 0.645, 0.741,0.862) and (1.397,1.037,1.551,1.349,1.160).
tolerance and VIF of (V, W ,X,Y and Z) with B are (0.716, 0.964, 0.645, 0.741,0.862) and (1.397,1.037,1.551,1.349,1.160).
tolerance and VIF of (V, W ,X,Y and Z) with C are (0.716, 0.964, 0.645, 0.741,0.862) and (1.397,1.037,1.551,1.349,1.160).
The five point likert scales are used to collect data for all the dependent and independent variables.
From the above result what types of interpretation can I make on A, B and C?
Surajit -
Thinking about continuous data, it sounds like you might be referring to multivariate, multiple regression. Multivariate because of the multiple y-values (dependent variables) of interest (A, B, and C), each with the same 'independent' multiple predictors. The collinearity would be with regard to the predictors in each case. It does not sound reasonable that the VIFs would be the same in each case. (I do not know what "tolerance" means here.) This is not familiar to me, and since you mentioned likert scales, which is also not my area, I may just not understand what you are doing, but my guess would be that you are using your software incorrectly. Perhaps data and/or commands are input incorrectly, or there is a software error. Your repetitive results seem to me to indicate such a problem. If you did not write the software, then I think you need to study it to see exactly what it is doing.
By the way, as I said, not my area, but for likert data, wouldn't you use Poisson regression? See
https://onlinecourses.science.psu.edu/stat504/node/165
and links at the bottom of those notes for a discussion, but not with regard to multivariate regression. I don't think that they mention such a thing.
Cheers - Jim
- Badges
- Science topic