RELATIONSHIP BETWEEN LOGARITHMIC SERIES MODEL AND OTHER SUPERPOPULATION MODELS USEFUL FOR MICRODATA DISCLOSURE RISK ASSESSMENT

抄録

Fisher's logarithmic series model (Fisher et al. (1943)) is a classical model in statistical ecology. In this paper we show that this model is a key model linking three models discussed in Takemura (1997), i.e., Poisson-gamma model (Bethlehem et al. (1990)), Dirichlet-multinomial model (Takemura (1997)), and Ewens model (Ewens (1990), Sibuya (1993)). This connection opens up the possibility of applying existing techniques of statistical ecology to the problem of microdata disclosure risk assessment.