When censored observations are present, the argument on p. 112 concerning the modified Wilcoxon test is fallacious. It is valid when tied observations are present provided that the definition of Vijk on p. 107 is modified by putting Vijk = i (Uij = 0 or Uik = 0). With this adjustment expression (2 8) on p. 107 is valid in the presence of ties but not in the presence of censoring. Valid expressions for Z and V in the presence of rightcensoring can be found from Prentice (1978) or from Kalbfleisch and Prentice (1980). In the latter reference equation (6 6) gives Z if used with (6 9) and equation (6 21) gives V. The vector z corresponding to each observation should take the scalar value 1 for members of the first sequence and 0 for members of the second in this special case.
No takes yet. Share an insight, caveat, or question.
Bellhouse et al. (1981) studied this question.