Statistical Analysis of Structure using Latent Variable Model

在本研究中，我们主要从经验建模方法、贝叶斯估计和模型比较的角度对潜变量模型进行了研究。利用马尔可夫链蒙特卡罗（MCMC）方法，这是基于模拟，我们开发了有效的估计方法和贝叶斯统计模型的使用。这是我们四年计划的最后一个阶段。我们计划通过总结每个研究者的研究和在该项目的五个国际研讨会上提交的研究论文来完成该项目。2006年在日本举办了两次国际研讨会，并与奥地利高等研究院（IAS）联合举办了一次研讨会。在9月的第一次研讨会上，我们邀请了五位著名的外国研究者，A。C.哈维，H。货车Dijk，A. C.卡梅隆，C. Meghir和J.M.罗宾我们还与年轻的研究人员进行了重要的交流， 关于我们 东京大学的优秀研究论文。这次会议是与日本统计协会成立75周年研讨会联合举办的，题目是“应用计量经济学的最新发展”。第二次会议“贝叶斯统计和应用计量经济学国际研讨会”于11月在东北大学举行。我们邀请了六位国际知名的计量经济学家。本次研讨会的主题是根据贝叶斯分层模型等主要使用的建模计算方法。Geweke，Chib，Poirier，Tsay，Jeriazkov，Tsurumi和其他日本研究人员共进行了21次演讲。在国外，我们共同主办的“JEuBES 2006，1^<st>日本-欧洲贝叶斯计量经济学和统计学会议”在维也纳的IAS举行。来自日本的5名小组成员和来自欧洲和美国的13名与会者参加了会议。研究课题是关于贝叶斯模型，层次结构和空间模型等，在6月，四个人参加了瓦伦西亚国际会议贝叶斯统计，并在8月一个参加了联合统计会议在美国举行，这将是必要的，考虑到一个共同的建模方法，许多类似的模型在更复杂的系统在未来。少

In this research, we investigate the latent variable models mainly from the view points of the empirical method of modeling, the Bayesian estimation, and the model comparison. Making use of the Markov chain Monte Carlo (MCMC) method, which is based on simulation, we developed efficient estimation methods and usage of Bayesian statistical models. This is the last period of our four-year-project. We planned to finalize the project by summarizing the each investigator's researches and the research papers submitted in the five international symposiums of this project. We will be searching for the direction of the researches for the future.In 2006, we held two international symposiums in Japan and a workshop jointly with IAS (Institute for Advanced Studies) of Austria. At the first symposium, in September, we invited five famous foreign researchers, A. C. Harvey, H. van Dijk, A. C. Cameron, C. Meghir, and J. M. Robin. We also had a significant exchange with young researchers who presenting … More excellent research papers at the University of Tokyo. This conference was held under the title of "Recent Development of the Applied Econometrics" jointly with Japan Statistical Association 75'th anniversary Symposium. The second meeting "International Workshop on Bayesian Statistics and Applied Econometrics" was held at Tohoku University in November. We invited six world-known econometricians from abroad. This workshop centered on the topics of the computational method according to the modeling mainly used by Bayes hierarchical model, etc. There were totally 21 presentations given by Geweke, Chib, Poirier, Tsay, Jeriazkov, Tsurumi, and other Japanese researchers. In abroad, "JEuBES 2006, 1^<st> Japanese-European Bayesian Econometrics and Statistics Meeting", which we co-sponsored, was held at IAS in Vienna. 5 group-members from Japan and 13 participants from Europe and the US joined the meeting. The research topics were about Bayes models, the hierarchical structures, and spatial models, etc. In June, four people attended the Valencia international meeting on Bayesian Statistics and in August one attended the Joint Statistical Meeting held at U.S.A. It will be necessary to take into account a common modeling method about many similar models in more complex systems in the future. Less