Abstract

Efficiency plays an important role in the assessment of a design. An efficient design utilizes the design structure and information effectively. This study comparatively examines the design efficiencies of D- and A-optimal designs for Poisson regression models with two and three variables. The optimal designs were constructed for the considered models and their design efficiencies evaluated. The two-variable Poisson regression model was observed to yield greater efficiencies than the Poisson regression model with three variables. The A-optimality criterion produced greater design efficiencies both for the two- and three-variable Poisson regression models than the D-optimality criterion. The study recommends that the two-variable Poisson regression model should be broadly preferred to the three-variable Poisson regression model especially for A-optimality.