A convenient approach to the statistical analysis of distributions for
counts is possible using the empirical probability generating functio
n. In this paper we give an overview of recent results and show the us
efulness and advantages of this methodology. On one hand, there are so
me stochastic models in which the probability generating function aris
es naturally and therefore it seems reasonable to use its empirical co
unterpart. On the other hand, this statistical tool has demonstrated t
o be useful in the study of classical statistical problems of distribu
tions for counts, especially in exploratory data analysis, rapid multi
-parameter estimation and testing the goodness of fit. Our recommendat
ion is to make allowance for the empirical probability generating func
tion when dealing with statistical inference for discrete distribution
s.