This article works through the basic theory of estimating the parameters of a complete economic model by full information maximum likelihood, and develops a simplified and condensed routine. The routine is then applied step by step to the estimation of the parameters of a known numerical model which has generated its own data. Various other methods of estimation are less costly, and some of these are tested both on the above synthetic data and on the actual data of a small Canadian model. The aim here is to attempt to make a rough efficiency appraisal, in terms of cost of estimation and bias, of the various methods tested.