Analysis of incomplete blocks for rankings

Each of b judges independently rank kh out of t objects, h = 1, 2, ... , b. This basic design is replicated n times, resulting in nb independent blocks. The methods introduced in Alvo and Cabilio (1991) to test for concordance among the judges in a BIBD are extended to this situation and applied to a statistic based on Kendall's tau. This statistic is then analyzed for a situation not considered in Alvo and Cabilio (1991), and it is shown to maintain its superior efficiency over the Durbin test.