Economical Multifactor within-Subject Design Robust against Trend and Carryover Effects

摘要

When the same subject is tested serially across the conditions of an experimental design, unless precautions are taken, there is a large chance that the experimental results will be biased by sequence effects. In human performance research, two types of sequency effects are quite common, i.e., trends through the data, such as learning (and even systematic variations in the equipment and/or environment) and carryover effects, when the characteristics of one condition influence the performance on the condition that follows. Unless these effects are removed of neutralized, the information obtained from the experiment will be distorted and incorrect conclusions may be drawn. Whether 2 sub k factors or 2 2 sub k-p fractional factorials are used with the within-subject approach, sequence effects must be properly handled. The traditional way of counterbalancing or assigning conditions at random can be inadequate and certainly uneconomical. When large-scale multifactor experiments are performed, concern with economy is paramount. In this report, a within-subject design constructed from 2 to the 5th powerfactorial plan is evolved into a 2 to the seven minus two power fractinoal factorial, Resolution IV, design.