Center for Automated Learning and Discovery
School of Computer Science, Carnegie Mellon University


On Limiting Random Partition Structure Derived
from the Conditional Inverse Gaussian-Poisson Distribution

Nobuaki Hoshino*

February 2002

Note: This report is provided in a draft format and
should not be cited without the consent of the author.


Keywords: Random clustering, species abundance, superpopulation, disclosure risk

In the present article, we derive a new multivariate distribution that belongs to an exponential family through a limiting argument over the conditional inverse Gaussian-Poisson distribution proposed by Hoshino. The derived distribution can be used as a model of random partitioning of positive integers, which is relevant to applications in many fields such as statistical ecology, linguistics and statistical disclosure control to name a few. We clarify some properties of this distribution that are important in applications.

17 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by