week 10: moderation

Question 1

What do tests of moderation and mediation allow us to understand?

Answer 1

why a relationship exists

Question 2

what does moderation identify?

Answer 2

• can help identify for who or under what conditions (when) a relationship between variables occurs
• where the nature between the variables is dependent on the level of the moderating variable
• attempts to identify causal condition(s) for when a relationship occurs

Question 3

what does mediation identify?

Answer 3

• explains the sequential order of variables
• can help explain how/why there is a relationship between variables
• attempts to identify causal pathways between variables

Question 4

what is the difference between moderation and mediation?

Answer 4

• moderation explains for who and when a relationship occurs. identifies causal conditions for when a relationship occurs
• mediation explains how and why a relationship occurs. identifies causal pathways between variables.

Question 5

What does the moderator variable affect in terms of variable relationship?

Answer 5

strength or direction of the relationship between the predictor and the outcome variable

Question 6

how did Baron & Kenny (1986) test for moderation?

Answer 6

1. transform predictors using grand mean centring
2. create a product (interaction) of the predictor and the moderator
3. put the (centred) predictor, the (centred) moderator and the interaction into the multiple regression analysis all as predictors
4. if the interaction is significant, moderation is present

Question 7

How do you transform predictors using grand mean centring?

Answer 7

• calculate the mean for X
• calculate the mean for W
• create a centred score for X (for each score on X, subtract the mean of X)
• create a centred score of W (for each score on W, subtract the mean of W)

Question 8

How do you create a product (interaction) of the predictor and moderator

Answer 8

multiply the two centred scores of X and W (predictor x moderator)

Question 9

Why do we centre the variables?

Answer 9

when you centre variables, you are left with each participants score minus the mean
- reduces multicollinearity between predictors and interaction term (multicollinearity = when predictors are highly correlated)
- the interaction term would be highly correlated with the predictor and moderator (because it is based on them)
- without centring, it would be difficult to separate and interpret the unique effects of each variable

Question 10

when running a moderation ‘PROCESS’ calculation in SPSS, what does ‘coeff’ on the output mean?

Answer 10

coeff = the unstandardized beta

Question 11

when running a moderation ‘PROCESS’ calculation in SPSS, what does ‘se’ on the output mean?

Answer 11

se = the standard error of the unstandardised beta

Question 12

when running a moderation ‘PROCESS’ calculation in SPSS, what does ‘LLCI’ and ‘ULCI’ on the output mean?

Answer 12

upper and lower confidence intervals

Question 13

in moderation, what is W, X, and Y in terms of variables?

Answer 13

W = moderator
X = predictor
Y = outcome