ABSTRACT

Like goodness of fit for suits, dresses or trousers, one curve may fit another well in some places and fit badly in others. For this reason, Neyman (1949, p. 259) discusses a modified version of the Poisson density; a curve that in its untruncated form equals e~x\x/x\ for x = 0 , 1 , 2 ,... and equals 0 else­ where. Consider, for example, a fictional population comprised of sultans who father 50 or more children. Suppose the variate of interest here is the number of sons within such a family who at some time in their lives are more than 6.3 feet tall. Then, because its value is determined by counting an exact number of rare events, variates like this are often taken to be “Poisson distributed.”