Resumen de Estimating vote party entries and exits by ecologicalinference. Mathematical programming versus Bayesianstatistics

Rafael Romero, José Manuel Pavía Miralles

• Ecological regression has been very fertile in proposing procedures that can be used to estimate the so-calledvote transfer matrices. According to various studies, the method implemented in theRfunctionei.MD.bayesis the one that currently presents the best performance. This method, based on a hierarchical Multinomial-Dirichlet model, is estimated from a Bayesian framework, which demands highly trained and experiencedusers. The procedure also seems to require having data for a large number of voting units to generate goodestimates, which increases both its computational and operative burden. In this work we show through anexample how the new algorithm available in theRfunctionlphom, based on linear programming, is highlycompetitive, generating solutions comparable to the ones attained withei.MD.bayeseven with few units,which drastically reduces its computational and operative costs.lphomis also very easy to use.