|
Abstract:
|
Many criteria exist for which design matrices can be compared for efficiency to optimality. Most standard design matrices for first- or second-order models are limited from improvement due to the fixed values associated with their construction. Central composite designs permit users to specify the axial values, thus allowing restricted optimization to be made of the design matrix. A method will be discussed for determining G-optimal central composite designs based on their efficiencies. These efficiencies will be compared to central composite designs incorporating face-centered, spherical, and rotatable axial values for a varying number of factors and replication of center runs. This same information will be reported for D-optimality, A-optimality, and E-optimality, as well as a new class of VIF-optimal designs, which measure efficiency relative to orthogonal design matrices.
|
