+ 1 ] 2 " : nN

When no misclassification has occurred and when thus 6 = 0, the first equation of (14.2.17) yields the well-known estimate

In the event that (i) all sample observations are k's, in which case rk = N, rk+ 1 = 0, and x = k, or (ii) all sample observations are (k + I)'s, in which case, rk = 0, rk+ 1 = N, and x = k + 1, then estimates p and 0 do not exist. In the event that rk = rk+ 1 = 0, no estimate is possible for 6, but from the first equation (14.2.15), we would have p = ~~=o xrxfnN. Although of theoretical interest, these exceptions seem unlikely to be important in practical applications

for the same reasons that were noted in connection with estimation from misclassified Poisson data.