Inference on trends in several Poisson or binomial populations

It is assumed that either the means of several Poisson, or the parameters of several binomial populations are linear functions of a regressor variable such as time. If samples from the different populations are given at several values of the regressor variable, the maximum likelihood solution exists for a reasonable set of hypotheses. It can be found by a necessarily converging, iterative procedure of successively solving single equations. Two examples are given, one dealing with suicide numbers in the different districts of Vienna over the years 1966–1974 and one with proportions of stillbirths among all births in different areas of Austria during the years 1951–1960; on the basis of these data three different solution algorithms are compared.