Key points are not available for this paper at this time.
Rank analysis of incomplete block designs using a method of paired comparisons is covered in the references. In these papers it is assumed that there is an experiment consisting of I treatments with n repetitions of each of the 1(1 1)/2 pairs (i, j). It is furthermore postulated that associated with each of the treatments, denoted by T1, I , T, ,there exist parameters, 7ir for Ti , such that 7ri > 0 and i=, 7ri1. The parameters are further defined with the probability statement that, if Xi generally denotes an observation on a sample of Ti,
Otto Dykstra (Wed,) studied this question.