Moderation Flashcards
What is moderation?
Investigates when X predicts X - does the relationship vary depending on the third variable
moderation = interactions
What does it mean?
The relationship between X and Y changes depending on the value of some other variable - the moderator
the strength / direction of the relationship is conditional on the value of another variable (the moderator)
What changes?
The relationship changes depending on the value of W
Example
Number of parties attends predicts self-esteem
but this relationship varies depending on extraversion
for extroverts - the relationship is positive
for introverts - the relationship is non significant
Modifies the relationship - X and Y might be strongly positively related when the moderator is high, but non significantly related when the moderator is low
How do you run the regression?
With 3 predictors
The predictor and moderator as individual predictors - main effects
The product of the two - interaction XW
How do we know if we have moderation?
If the interaction is significant
Why do you need to include the main effects for the predictor and the moderator?
Otherwise the interaction and main effects are confounded and a significant interaction can’t be interpreted
At what value of the moderator do we want to estimate the coefficient for the predictor, when there is no interaction?
The coefficient for the predictor changes depending on the value of the moderator, b1 = effect of a predictor on the outcome, when all other variables in the model are 0. When there is no interaction, this doest matter because the b parameter doesn’t change
When there in an interaction, what value of other predictors do we want to estimate the b parameter?
The b parameter changes depending on the value of the other predictors (moderator)
zero doesn’t usually make sense, e.g. heart rate, response time, height etc
The mean is the most sensible option
Solution: transform the variables so that it has a mean of zero
How do you grand mean centre?
Simply subtract the mean from each participants score - for the predictor and the moderator - not the interaction
each persons score then becomes the deviation form the mean - makes the output more interpretative
What happens in grand mean centring?
The interpretation of the predictor’s b parameter becomes: at the mean of the moderator, there was a significant positive relationship between the predictor and the outcome. At different values of the moderator, the relationship might be different
How can we interpret a significant interaction?
Using a simple slope analysis - tells us what the coefficient for the predictor is at different values of the moderator tells us the relationship: at the mean of the moderator when moderator is - 1SD below the mean when moderator is + 1SD above the mean look on conditions effects
What is a simple slope?
Probing interactions by looking at the X Y relationships at different values of the moderator
What does moderation then show for the example?
The relationship between number of parties and self-esteem is conditional upon a person’s extroversion score.
When extroversion is high (+1SD from mean), number of parties is strongly positively related to self-esteem
At the mean of extroversion (0), number of parties is positively related to self-esteem
When extroversion is low (-1SD), number of parties is non-significantly related to self-esteem
When can you use moderation?
Anything you can analyse with regression you can do
Survery research - self-report questionnaires. Need a theory first, which variable depends on your theory
Experimental research - investigates when your experimental manipulation influences the DV
