Too Many Questions? The Uses of Incomplete Cyclic Designs for Paired Comparisons

Abstract

The author describes a method for constructing incomplete designs for paired comparisons data, using cyclic designs that place each item in the same number of pairings. He tests the reliability of the method using judged similarity data on twenty-one animal names and finds that designs that use 30% of the data have an acceptable level of reliability. The method has great potential value in field research as a means of reducing the number of questions required.