Improvement of the Sampling Methods in Official Statistics and Assessment of the Disclosure Risk of Microdata Sets

我们的目的是评估微观数据的披露风险，并改进官方统计的抽样方法。观察到的样本量指数和估计的群体规模指数经常被用来评估从群体中抽样的微数据的披露风险。在基于样本量指标的总体规模指数估计中，采用超总体模型的参数化方法是目前的主流方法。提出了一种非参数极大似然估计方法，该方法对人口规模指标进行了一些简单的限制，并尝试采用广义Zipf模型进行限制。在对微数据进行风险评估的同时，还探讨了利用亚种群信息进行风险评估的方法。对屏蔽微数据进行了研究。我们被允许使用三个官方统计数据，劳动力调查，家庭收入和支出调查和全国家庭收入和支出调查，然后我们可以为每个统计数据提供安全的微数据集的形式。在检视劳动人口调查的时间序列时，我们考虑了另外两个统计数字，除了使用住户的属性外，也使用了个人的属性。我们每年与其他研究小组联合举办一次研讨会和几次研讨会，与各领域的研究人员交流思想。我们策划了《统计数学研究所学报》的特刊《微数据披露的统计理论》，收录了我们的文章。我们的研究成果也发表在各种期刊上。

Our aim was to assess the disclosure risk of microdata and to improve the sampling methods of the official statistics.The observed sample size indexes and the estimated population size indexes are often used to assess the disclosure risk of microdata which is sampled from a population. In estimation of population size indexes based on sample size indexes, the parametric method with a superpopulation model is the mainstream today. We proposed a nonparametric maximum likelihood estimation method in which some simple restrictions for population size indexes were used, and tried to adopt the generalized Zipf's model for the restriction. The estimation method using information of subpopulations was also investigated Simultaneously with the risk assessment of microdata, some problems for the tabular data and usefulness of the. masked microdata were studied.We were permitted to use three official statistics, Labour Force Survey, Family Income and Expenditure Survey and National Survey of Family Income and Expenditure, and then we could present the form of the safe microdata set for each statistics. In examination for Labour Force Survey the time series were taken into consideration, and we used not only attributes of households but also those of individuals for other two statistics.We held one symposium and several workshops every year in alliance with other research groups to exchange ideas with the researchers in various fields. We planed the special issue "Statistical Theory for Disclosure of Microdata" in Proceedings of the Institute of Statistical Mathematics, in which our articles were contained. Our research results were also presented in various journals.