Hey, just like Book et al. (2015) did in their article (Article Unpacking Evil: Claiming the Core of the Dark Triad
) I would like to compare two models of canonical correlations, whilst one is nested in the other. Just like Book et al. I would like to see, if the addition of some variables maximize the explained variance (1-Wilks Lambda = R²).I'll conduct my analysis in R.
Greetings
Hello Maximilian,
It appears that authors of the Unpacking evil... study simply treated each canonical correlation as independent of the others. As each used the same set of DVs ("dark triad"), with the same sample in each of the two studies, the correlations are dependent, not independent.
What one could do is: (a) incorporate all IV sets into an omnibus model, then (b) successively omit one set (e.g., Big 5) from a subsequent model, and compare the change in canonical R, as nested models.
The real question is, does it make more sense to evaluate each IV set individually (as done in the article), or approach the framework as SEM with all IV sets (as implied by the previous paragraph)? Within the SEM approach, individual effects (collective aggregates of a given IV set) could be posited as influences on the DV set, and directly compared for relative strength.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Books