The structure of cyclic paired-comparison designs

Abstract

When n objects are to be compared in pairs, a complete experiment requires N= comparisons. There are frequent occasions when it si desirable to make only a fraction F of the possible comparison, either because N is large or because even an individual comparison is laborious. The problem of what constitutes a satisfactory subset of the comparisons has been considered by Kendall [5[ who lays down the following two minimum requirements: (a) every object should appear equally often; (b) the design should be ‘connected‘ so that it is impossible to split the objects into two sets with no comparison made between objects in one set and objects in the other.