Abstract.

A quasi-experimental design with pre and post- test was used to estimate training effects for the University of Costa Rica’s admission test, a standardized exam that measures reasoning abilities in mathematical and verbal contexts. Four secondary public schools from the metropolitan central area of the country participated in the study; two of them were randomly assigned to the intervention group and the other two to the control group, with 61 students in the first group and 80 in the second. The intervention consisted of five three-hour training sessions, using a written guide developed by a pedagogy expert with a constructivist approach. Before and after measures were reduced test forms of the real admission test from the year 2014. The dependent variable was the difference between the two measures. The effect of the training was estimated using a multilevel Bayesian regression model with a significant magnitude of 3.5 percentage points.

Keywords: Impact evaluation; Bayesian hierarchical models; effects of training; university admission test; multilevel models