For simplicity let us consider a single output Y and a single endogenous regressor of interest S. Suppose we have run multiple experiments over different periods in time, where we in each experiment, we have multiple treatment conditions which produce different amounts of variation in S which results in changes in Y only through S (exclusion restriction). You could imagine that we have exogenous covariates X, recorded for each experiment for every unit in the experiment. Let us assume a constant causal effect p from S -> Y and let us assume that this does not vary with time. We can run a TSLS for each experiment to get an estimate for p. My question is what is the best way to combine all these experiments to produce a single "best" causal estimate for p.

  • $\begingroup$ I would consider mvmeta in Stata or R. $\endgroup$ – Joe_74 Feb 18 '19 at 10:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.