On The Choice of a Completely Randomized Design with Emphasis On The D-Optimality Criterion

Abstract

Completely randomized designs (CRDs) are designs in which treatments are allotted to the experimental units by chance. The designs include the unconstrained CRD, the zero-sum constraint CRD and the baseline constraint CRD. Since there are several CRDs, we are faced with choosing a design that will yield the optimal result. In this paper, we present the use of the D-optimality criterion as an objective tool in making such a choice. The study shows that the baseline constraint CRD is the most suitable choice of designs for an experiment with five treatments and six replicates.

Keywords: Experimental design, CRD, D-optimality criterion, treatment, replicate, model, regression, information matrix, measure, eigenvalue, determinant.