Research on statistical analysis of evaluating fixed-assets for housing and its application

这个项目的主要贡献是发展评估住房固定资产的统计模型，并将发展的统计方法应用于某些计算机应用。在模型开发部分，我们基于资格理论一对现有模型进行了细化，将其扩展为多元回归模型。这就产生了基于多元回归模型的评价方法。除了这些确定性方法外，我们还重新考虑了Cox回归在固定资产评估中的适用性。该方案采用协变量选择方法，使定价更加准确。此外，我们考虑了缺失数据问题，并分析了多重分类分析下协变量之间的关系。实现上述统计方法的系统已分别构建为电子表格插件和基于web的应用程序。(1)基于多元回归模型的评价研究更多：回归模型一般是对资格理论i的泛化模型，因此将基于资格的评价方法扩展到多元回归模型中，这样我们就可以将不平移的站点距离等定量数据处理为离散值，即分类。此外，基于多元回归的精细化模型导致了多重分类分析中的主成分分析。（ii）Cox回归模型：将现有方法扩展为Cox回归方法。这个扩展使得协变量的选择成为可能。此外，我们还开发了EM算法来处理Cox回归模型中的缺失数据。（三）系统开发：为了将上述提出的方法应用到实际的固定资产评估领域，我们构建了两类有用的应用程序。首先是电子表格的插件软件，便于所有用户使用。第二个是基于web的应用程序，它由一些数据管理体系结构组成。少

The main contribution of this project was to develop statistical models for evaluating fixed-assets for housing and to implement of the developed statistical methods into some computer applications. In the part of development of models, we refined the existing model based on qualification theory I, so that they were extended to the multiple regression models. This led to the evaluation method based on the multiple regression models. In addition to such deterministic methods, we reconsidered the applicability of Cox regression to the fixed-assets evaluation. The covariate selection method was applied to make the pricing accurate in this scheme. Moreover, we made consideration on missing data problem, and analyzed relationship between the covariates under the multiple classification analysis. The systems which implement the above statistical methods have been built as a spreadsheet plug-in and Web-based application, respectively.(i)Study on evaluation based on the multiple regression mod … More el : The regression model is generally an extensive model to the qualification theory I. Thus the qualification-based evaluation method was extended to the multiple regression model, and therefore we can treat quantitative data like distance from a station without translation to discrete values, namely classification. In addition, the refined model based the multiple regression led to the principal component analysis in multiple classification analysis.(ii)Cox regression model : The existing methods were extended to Cox regression methods. This extension made the selection of covariates possible. Also, we developed EM algorithm to deal with the missing data in the Cox regression model.(iii)Development of systems : In order to distribute the above proposed methods into the area of practical evaluation of fixed-assets, we built two kinds of useful applications. The first is a plug-in software of spreadsheet, so that it is easy to use for all users. The second is a Web-based application, which consists of some data management architectures. Less