Center for Automated Learning and Discovery
School of Computer Science, Carnegie Mellon University
On a Conditional Inverse Gaussian-Poisson Distribution
Note: This report is provided in a draft format and
The present article treats a distribution of random partitioning of the positive integer. Although such a distribution is important concerning applications in many fields such as statistical ecology, linguistics and statistical disclosure control, not very many models are known owing to the difficulty caused by inevitable combinatorics. The present article shows that conditioning the total frequency of an inverse Gaussian-Poisson population model leads to a new manipulatable distribution of random clustering. We also give formulae that are necessary for the application of this distribution.
*Visiting faculty from the Faculty of Economics, Kanazawa University,