On goodness of fit tests for the Poisson, negative binomial and binomial distributions (Q2633412)

For an independent sample with discrete values, this paper considers the classical goodness-of-fit problem to test whether the distribution of the random sample is in a family of discrete distributions with an unknown parameter. Three specific distributions, Poisson, negative binomial and binomial, are considered in this paper. Although there have been many goodness-of-fit tests proposed for these distributions, two new tests are proposed in this paper based on some approaches proposed in recent literature. The first test is based on a likelihood ratio test proposed by \textit{J. Zhang} [J. R. Stat. Soc., Ser. B, Stat. Methodol. 64, No. 2, 281--294 (2002; Zbl 1067.62046)] for the skew-normal distribution, which was shown to be more powerful than traditional tests. Another test is based on an approach which uses the conditional probability given a sufficient statistic and has an advantage that no approximation to the distribution of the test is required. Simulations are performed to compare the powers of the proposed tests with that of some traditional tests and a recommendation for the use of the proposed tests is made based on the results of the simulations.