Abstract: It is assumed that a moderately large sample is
available from a finite mixture distribution, the components being discrete
distributions for which moments exist. For s
components there will be at least parameters. Psi-function
differences create a random variable which itself leads to a test function
statistic, this statistic in general being asymptotically normal. Eleven
percentage points to the distribution of the standardized test function
(location and scale free) are given.
Keywords and phrases: goodness of fits test, logarithmic derivatives, maximum likelihood estimator, moment estimator, Psi-function.
