Key points are not available for this paper at this time.
Remarks. The fact that the categories with which the probabili-ties are associated is not known clearly causes a loss of some information as compared to the case in which the association is known. An indication of the loss of discrimination information (Kullback, 1959) for discriminating between the nonuniform case leadiiig to (2.1) and the uniform case, as a consequence of the grouping of 100 categories when n = 100 into the 'li sample configurations as in Table 2.1 may be obtained by computing the difference between (Kullback, 1959. p. 111)
Mantel et al. (Sun,) studied this question.