A- and D-optimal designs in real life data using imperialist comparative algorithm

Abstract

In this paper, we used Imperialist Competitive Algorithm (ICA) to get A-and D-optimal de­signs for two models; one with nonlinear and the other linear model were found for two life data. It was noted that the number of iterations for D-optimal design for both models was smaller than that of A-optimal design. This implies that getting a D-optimal design is preferable and saves time and energy compared to A-optimal since they are addressing the same issues related to the variance-covariance function. The Efficiency Lower Bound and the Sensitivity function embedded in ICA which support the general equivalent theorem in getting optimal design were also found in this study. Another advantage of using ICA algorithm is that it does not require the design space or the region of uncertainty to be discretized which means that the search for the support points of the optimal design is not restricted to the grid point. This was also confirmed here that the optimal design for A and D- designs were not restricted to the grid points.