← Other publications

Monte Carlo experiments test models on artificial datasets with known properties to assess the likely performance of an estimator in empirical work. Although the increasing savvy of political methodology has brought more and better Monte Carlo work, it is not always presented clearly or thoroughly. In particular, Monte Carlo results often appear in unwieldy tables rather than elegant graphics. I propose five guidelines for Monte Carlo graphics. I also define five graphic styles which help show the comparative performance of models over the parameter space, even when the models and parameters are many.