Mediation and indirect effects Flashcards
What are two research questions SMR can address (with corresponding statistics)?
- How much variance can the set of predictors explain together? (R2, model F)
- What is the relative importance of each predictor? (b, B, sr2)
What is a mediator?
A third variable, outside of the predictor and criterion, which explains or accounts for the relationship between IV and DV.
Why is it important to understand indirect relationships? (2)
To understand why something works.
If there are multiple mechanisms, we want to understand the important ones.
How many relationships are considered in a mediation analysis?
four
What contribution can we expect and not expect a mediating variable to make within a mediation analysis? Why?
Significant coefficient (beta) but no change for R2 expected/needed. The mediator is supposed to explain the IV-DV relationship, and does not need to explain additional variance.
What shows full and partial mediation?
The prediction coefficient of the IV on the DV is no longer significant when the mediator is considered in the model. If its still significant this is partial mediation.
What does the Sobel test ask?
Is the indirect effect reliable?
What are 5 conditions for testing mediation with multiple regression?
- IV predicts mediation
- IV predicts DV in block 1
- Mediator predicts DV in block 2 (I.e. when IV is controlled for)
- Coefficient for IV decreases to n.s. (full mediation) or decreases but still sig (partial)
- Sobel test shows indirect effect is reliable
In mediation, when is an SMR used? what is the alternative?
If there is more than one IV or control variables included. otherwise, correlation matrix can suffice for showing relationship between IV and Mediator
In mediation, what if IV doesn’t predict DV (in block 1) before considering Mediator?
Refer to it as an indirect effect of IV instead of ‘mediation’.
What is a suppression model?
When the IV has relationships with two mediating variables, one positive and the other negative. This give the impression of suppressing any effects of the IV on DV, through opposing action via the two pathways.
Which Betas and changes in Betas are reported for variables in HMR?
Report betas from entry block for each variable. (Change in betas between blocks is not considered.)
Which Betas and changes in Betas are reported for variables in Moderated Multiple Regression?
Report betas from entry block for each variable. (Beta change between blocks is not considered.) If interaction is significant, report beta for IV from final block of simple slope analyses.
Which Betas and changes in Betas are reported for variables in Mediation analyses?
Report betas from entry block and change in IV when mediator is entered.
