I have been trying to understand 2 stage least squares regression, and I am having difficulty understanding how that is different from mediation analyses. It would be helpful if you could provide some context, for instance, what sort of questions or hypothesis are these 2 techniques tests and how do they differ?
In 2SLS (i.e. instrumental variable regression), you want to find the unbiased effect of x1 on y. In order to to this, you need a variable, z (i.e. the instrument), that affects x1 but does not effect y (other than through x). In mediation analysis, you specifically want to find how x1 affects y indirectly via another variable, x2, possibly in addition to x1's direct effect on y.
Very rightly said by Christer Thrane. In mediation analysis, a mediating variable (M) is meant to increase the effect of X on Y thereby enhancing an existing direct relationship.
There has already been a discussion on this subject on ResearchGate. Look at https://www.researchgate.net/post/How-to-test-for-causal-mediation-with-instrument-for-mediator
A Google search of "2SLS versus mediation analysis" revealed a 2016 presentation by Graham Dunn and Richard Emsley on a close subject.
Good luck.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression