We consider the discrimination between more
than two hypotheses by means of the
Sobel-Wald-test for the parameter of a discrete
distribution belonging to the one-dimensional
exponential family. Based on an exact method
for the computation of the characteristics of
the underlying sequential probability ratio
tests we present a special discrete procedure
for the exact computation of the average
sample number function of the
Sobel-Wald-procedure. Examples are presented
for the Bernoulli and Poisson distribution.