4: Path Models and Mediation Flashcards
What is the implied correlation in a path model?
The sum of compound paths.
What are direct effects?
A single, forward facing arrow.
What are indirect effects?
A sequence of forward facing, single headed arrows through mediators.
What is another name for indirect effects?
Composite pathways.
What are spurious effects?
Sequences (with backward arrows) where variables share a common cause which makes them appear to be related.
What are unanalysed effects?
Page involving a curved arrow that show a prior common cause represented outside the model.
What is the total effect?
The combination of direct and all indirect pathways.
What are under-identified models?
There are more parameters than observer information, so the model can not be uniquely estimated.
What are just-identified models?
Numbers of observed information matches the number of parameters, so the model has a unique solution.
What are over-identified models?
There are less parameters than observed relationships, so the model and the quality of the solution can be estimated.
What are the 2 forms of error in path models?
The difference between a prediction and a score, and the difference between implied and actual correlation matrices.
How are path models estimated?
By coming up with values for the path coefficients that minimise the difference between the implied and observer correlations.
What is step 1 of classical mediation?
Establish that X -> Y and estimate c using regression.
What is step 2 of classical mediation?
Establish that X -> M and estimate a using regression.
What is step 3 of classical mediation?
Establish that the mediator impacts in the DV while controlling for the IV and estimate b while controlling for X.
What is step 4 of classical mediation?
If c’ becomes 0 (or non sig.) then M completely mediates the relationship between X and Y.
Why is step 1 of the classic model not always necessary?
Mediation can occur eve if the zero-order correlation between X and Y is small.
What are 2 limits of the classic model?
Too little focus in trying to understand the implied model and it gets too complicated even there are more than 3 variables.
How does the PROCESS approach estimate the parameters of a model?
Using a series of regressions.
How does the PROCESS approach estimate the significance of indirect effects?
Using bootstrapping methods.
What type of statistical errors can collinearity lead to?
Type I and II errors.
Give 3 distortions collinearity can lead to.
Amplified parameters, larger standard errors, and far apart betas.
When do suppression effects occur?
When the greater part of collinearity is irrelevant to the outcome/DV.
What is suppression, in terms of shared irrelevant information?
The suppressor shares more with the other precursor than the DV.
What happens in negative suppression?
One of the path weights reversed direction relative to the correlation.
Define classic suppression.
The suppressor has 0 correlation with the DV but correlates with the IV, controlling irrelevant variance and increasing prediction.
How do we most often see suppression?
As a special case of collinearity when trying to “control” highly correlated predictors.
