Stata module to perform ‘fully interacted linear matching’, that is a fully interacted linear regression model in which the treatment dummy is interacted with each one of the other regressors. The overall effect – averaged for the treated and optionally also for the non-treated and the population – is then displayed, together with analytical standard errors.

In an evaluation framework this amounts to allowing the treatment effect to vary according to all observable variables. An F-test of no heterogeneous effects is also displayed.

film allows the user to perform various steps to increase the flexibility of the regression model until it resembles a matching estimator (see psmatch2) – from simple dummy variable OLS regression, to full interactions, to imposition of the common support – while retaining analytical standard errors and related saving in bootstrapping time.

It further allows for robust standard errors as well as for clustering.

Download ado file

Download help