A Computer Aided Design Strategy Incorporating Two-Level Fractional Factorials

Abstract

Finding suitable subjects for a design can be difficult when subject attributes, coded 0, 1, are treated as independent variables. A design strategy is presented which requires: 1) a historical data base describing each subject by a 0, 1 k-tuple, 2) a ranking of the subject attributes in order of perceived (a priori) importance, and 3) a computerized search algorithm that determines the maximum number of attributes which can be simultaneously considered in a two-level orthogonal fractional factorial given the ranking of attributes and the listing of available subjects. The algorithm can solve many practical problems in reasonable computation time. An application is demonstrated.