Discussion of J. Heckman's Lecture : the Principles Underlying Evaluation Estimators with an Application to Matching
The paper of James Heckman is an extremely interesting survey of a huge literature on evaluation estimators with a special attention to matching estimation. The article essentially concentrates on models with two levels of treatment (1 for treated and 0 for non treated). My discussion will consider a slightly larger case with a possible continuous ordered set of treatment levels.