fc9b003bb003a298c2ad0d05e4342bdc-AuthorFeedback.pdf

Depending onwhich underlying population generated the original data, we will get different counterfactual24 conclusions. In some cases, this may not be the global optimum. Unfortunately, the MIP to compute the global27 optimum has|D|2 boolean variables so computing the exact solution is infeasible for instances beyond 10-20 data28 points.