Mediation & Indirect Effects Flashcards
What test do you do if you want to know if A (manipulated) affects B
ANOVA or Regression
What test do you do if you want to know if C changes the effect of A on B
Interaction effect of A x C in ANOVA
What test do you do if you want to know if A (measured) predicts B
Regression coefficient and significance
What test do you do if you want to know if A predicts B when controlling for C, D and E
Regression coefficient & significance in Multiple Regression
ANCOVA
What is Moderation and what is Mediation
Moderation - addresses the question of when an IV affects a DV, otherwise known as interaction effects
Mediation - addresses the question of how an IV affects a DV (or why), if the link between and IV and DV is not direct then there are 2+ steps between the IV and DV involving mediators (mediating variables) - a more complicated example in image. note that it can also be done with more than one predictor/IV at the same time
How to test a mediation hypothesis
Refer to image
1 - regress DV on IV (path c)
2 - regress mediator on IV (path a) to determine if IV significantly predicts the DV and the mediator
3 - regress DV on mediator and IV simultaneously (paths b and c) to determine if the mediator significantly predicts the DV, and what happens to path c when the mediator is included (is the effect of IV reduced to non-significance?)
if YES the IV significantly predicts the DV and the mediator, and the mediator significantly predicts the DV, and the inclusion of the mediator reduces the IV to non-significance - then this informally indicates mediation
4 - test the significance of the indirect/mediated path using the Sobel test (use quantpsy sobel calculator( and if Sobel test indicates that the indirect path is significant then it’s consistent with a mediation hypothesis
NOTE - significant Sobel test doesn’t ‘prove’ or ‘show’ that mediation has occured, it can’t establish causal relations, only test whether data are consistent with the causal hypothesis
This is the output of a mediation using regression - what does it tell us?
What is the indirect effect calculation?
The first table tells us that attitude significantly predicts behaviour (path c - refer to image)
The second tells us the attitude significantly predicts intentions (path a)
The third table (testing attitude’s effect on behaviour when intention is added) tells us that attitude no longer significantly predicts behaviour (path c) and intentions do predict behaviour (path b)
Indirect effect is measured by path a coefficient * path b coefficient so in this instance it’s 1.627*0.406 this means b = 0.66
With confidence intervals - what makes the effect significant?
when 95% confidence interval lower limit is .0254 and upper limit is .3456 - is it significant?
when lower limit is -0.0115 and upper is 0.4534
If the value of 0 falls outside the interval then it is significant
YES - value of 0 falls outside interval
NO - value of 0 falls inside interval
How to do mediation in Jamovi without regression
GLM Mediation Model (use the jAMM module) - put DV in dependent variable box, the predictor/IV in the covariate box, and the mediating variable in the mediators box (see image)
Make sure to have Bootstrap (BC) selected as the confidence intervals (with an interval of 95% and a bootstrap rep of 5000)
Have IE components and beta ticked under ‘display in tables’
Have suggested paths ticked under ‘path model’
This is the output of a GLM mediation analysis - what does it tell us?
Give the summary calculations/findings
Component: Attitude -> Intention is path a (effect of IV on mediator)
Intention -> Behaviour is path b (from mediator to DV)
Direct: Attitude -> Behaviour is path c (from IV to DV, controlling for the mediator)
The table gives us the coefficient for the indirect (mediated) effect in the first row. 0 falls outside the upper and lower limits so the effect is significant
Indirect effect: b=0.66, se=0.15, 95% CIs=.392, .990
How would you report the summarising statistics of a GLM mediation output if there had been multiple mediators (example healthy eating intentions AND awareness of opportunities were mediators between attitudes to healthy eating and healthy eating behaviour)
Indirect path via intentions: b=, se=, 95% CIs=
Indirect path via opportunities: b=, se=, 95% CIs=
possible for one to be significant and for the other not to be
