Standard & hierarchical multiple regression Flashcards
How is shared variance dealt with in SMR? (2)
- It is included in tests of the overall model.
2. Excluded from tests of individual contributions (betas).
What is partialled out in partial correlation?
Variance from other predictors is removed from BOTH the IV and DV
What specific variance is used to test the individual contribution of a predictor/beta?
Partial correlation between predictor and DV
How is the overall model dealt with in both ANOVA and MR?
ANOVA: no test of overall model
MR: tests overall model automatically
How are the effects of IVs dealt with ANOVA and MR?
ANOVA: main effect of IV is tested, regardless of other variables’ effects/contributions (akin to bivariate correlation)
MR: tests unique effect of IVs, all other IVs’ effects are controlled/partialled out
Why, historically, does ANOVA assume IVs are uncorrelated? When does this become a problem?
First created for analysing experiments where random assignment meant IVs were not typically correlated. When analysing data without random assignment (e.g. blocking or natural variables)
How are interactions dealt with in ANOVA and MR?
ANOVA: tests all interactions automatically
MR: need to actively create an interaction term and test it with a hierarchical model
What does a multiple correlation coefficient represent?
A bivariate correlation between the criterion and the best linear combination (composite) of predictors
How do you calculate shared variance?
Model R squared minus the sum of unique contributions for all predictors. What remains is overlapping/shared variance in the model.
Why can’t we compare betas/IVs between different studies?
Betas rely on standard deviations which can differ between samples
What happens to shared variance in hierarchical regression?
It is attributed to IVs entered earlier in the model and informs the test of those IVs, making them more likely to be significant.
Why use a hierarchical regression model over a standard regression?
It allows us to give the shared variance to variables which theory tells us have a strong relationship with the DV. Otherwise, in an SMR the shared variance is never used in any tests of IVs’ individual contributions.
