Linear Regression Models: Mediation and Moderation Analysis Flashcards
What is meant by moderation?
A moderator variable influences the nature (e.g., magnitude and/or direction) of the effect of an antecedent on an outcome.
Moderation refers to the conditions under which an effect varies in size
The moderator changes the effect of the IV on the DV.
What is meant by mediation?
Mediation points to the presence of an intervening variable that transmits the effect of an antecedent variable on an outcome.
A mediator variable transmits the effect of the antecedent on the outcome, either in part or whole
IV affect DV through another variable (mediator)
What is the difference of partial and full mediation?
With full mediation thereis only an indirect relationship between IV and DV through the mediator.
In partial there is both a direct and indirect effect.
How to test for mediation?
Old school way:
- Run three regression:
Firs test the IV and DV - must be significant
second test the mediator and IV - must be significant
Third test all three - must be significant, and the coefficient of IV must be smaller here, then testing without the mediator.
Can also use boostrapping
How to evaluate the mediator?
Use the Sobel test for the old school way of testing the mediator.
It calculates the sig. of the indirect effect
It works well for big samples
Also check R^2 for the regression with and without moderator. R^2 for mediator=R^2 both - R^2 IV
R^2 increase if mediator help explain variation in DV.
What are some potential problems of mediators?
If the mediator effect is the opposite sign, the mediator off sets the direct effect and will show up insignificant.
The sobel-test is not reliable (always report bootstrapped result)
Measurement error especially due to common method variance bias.
How to test for moderation?
We are interested in the effect of the IV at different levels of the moderator. Hence, the interaction between IV and moderator.
Add both the IV and moderator plus the interaction in the regression (all at once, otherwise it cannot be interpreted)
Simple slopes analysis: mapping out the moderation by working out the values of DV as predicted by IV at different levels of the moderator
Is common method variance a problem with moderation?
No
What are the potential problems of moderation?
The measurement error of two constructs is higher when interacting them.
Balanced sample size, representing each category equally is required to ensure the statistical power.
Doing mean centering for binary variables
How do you evaluate the moderator?
R^2
We hypothesis that only employees with high self-control can benefit from autonomy, unlike those with low self-control who actually perform worse with higher autonomy - What is the mediator/moderator?
self control (Moderator)
Autonomy has better effects if the task is highly complex - What is the mediator/moderator?
task complexity (Moderator)
Autonomy increases performance through higher feelings of responsibility, that increases performance of the employee - What is the mediator/moderator?
feeling of responsibility (Mediator)
Employees with higher autonomy perform better because they can control their own schedule - What is the mediator/moderator?
control of time (Mediator)
