Moderation & Mediation Flashcards
What is moderation?
Moderation occurs when the relationship between two variables depends on a third variable.
When the two predictors interact leading to a mechanism being created.
Above the main effects.
What is needed for moderation?
Outcome must be continuous + normally distributed.
Continuous predictor + centred.
Why do we centre predictors?
Usually the interaction effect is interpreted when all other predictors are 0.
0 is often not sensible (sometimes not even defined in the scale!).
Centring - people with average value in both predictors.
How many effects will there be in moderation?
3. Main effect of X. Main effect of M. Interaction effect. (Basically like a multiple regression with three predictors).
What other program do you have to use for a moderation?
PROCESS.
How do you run assumption checks?
You can’t in PROCESS!
Give 3 advantages of PROCESS?
Centring automated, syntax to plot simple slopes, Johnson-Neyman method.
Explain the Johnson-Neyman method.
Need to look at the column of p-values and find where it becomes non-significant (if it changes!).
How do you compute a simple slope plot?
Need to copy the syntax into SPSS.
What is the definition of mediation?
The relationship between predictor and outcome variable can be partially explained by their relationship with a third variable.
Predictor influences the mediator, the mediator influences the outcome variable.
Indirect effect between X and Y. Mediated by M.
What is needed for mediation?
Outcome variables must be continuous.
Must use bootstrap method.
What is a, b, c’, a*b?
a - X-M. b - M-Y. c' - X-Y. a*b - X-M-Y. (indirect). c = c' + a*b.
What is the bootstrap method?
Allows us to create new datasets based on the one dataset we have.
Non-parametric method.
Why do we use bootstrap (mediation)?
To calculate standard errors properly (otherwise we can’t know if the indirect effect is normally distributed).
What model number do we use for mediation?
4.
Number of bootstrap samples - 5000.
What model number of we use for moderation?
1.
How do we check for the model assumptions in mediation?
PROCESS does not allow us to check.
But bootstrap is non-parametric anyway so this is not an issue!
How can we run a mediation model in SPSS?
It is impossible!
Only in SEM software - e.g. R.
What doesn’t PROCESS give us in indirect effects (mediation)?
A p-value.
Need to look at CIs.
What is an issue with mediation?
In principle - we assume a process over time!
Most study designs only care about the temporal assumptions.
How many effects do we have in a mediation?
3 direct + 1 indirect.
How can you interpret a moderation as an interaction?
The covariation of two predictors predict an outcome variable above and beyond the main effects of those two predictors.
Why do we use R-squared when we should use Adjusted R-squared?
Because PROCESS is still under construction and does not have the Adjusted R-squared yet.
What is a full mediation?
Link between X and Y is non-significant. (c’).
Indirect effect is significant.
What is a partial mediation?
We have a significant indirect effect but c’ is still significant.
Indirect effect is significant.
What is the difference in the interaction between moderation and mediation?
Moderation is between two predictors (X1 * X2).
Mediation is based on an interaction between two slopes (a * b).
Name one difference in the method between moderation and mediation?
Moderation - continuous predictors.
Mediation - bootstrap.
