Experimental Designs for Investigating Complex Human Operator/Machine Systems

Abstract

To increase the generalizability of her results, a human factors specialist must consider a large number of factors simultaneously when investigating complex human operator/machine systems. When complex multifactor experiments are necessary, the resulting number of treatment conditions and cost of conducting the study quickly becomes unwieldy if traditional, completely-crossed, factorial designs are used. Several data reduction designs are reviewed as potential alternatives to solve the generalizability/cost dilemma. These alternatives include single observation factorial designs, hierarchical designs, blocking designs, fractional factorial designs, and central-composite designs. Each of these alternatives should be part of a clearly formulated research strategy in which the experimenter efficiently collects her data in stages, completes a thorough and careful pretesting, determines the real-world constraints dictated by the research problem, and selects the necessary design modifications based on these real-world constraints.