Research on Integrated Time Series Analysis Softwares

通过前几个财政年度的研究，我们开发了一种针对一般状态空间模型的递归滤波和平滑算法，该算法可应用于时间序列建模的各个领域。基于这些结果，在这一年里，我们对开发的程序在各种实际数据集上进行了测试，并进行了必要的修改。我们还研究开发新的模型和更有效的算法。本研究基于非高斯状态空间模型，建立了一种分析具有非线性、非平稳性、非高斯性、异常值和缺失观测等多种特征的时间序列的统一方法。我们还对时间序列分析结果的图形化显示进行了研究，并为此开发了各种程序。特别是，通过使用CALCOMP图形库，可以在包括大型机和个人计算机在内的各种计算机上使用图形程序。如上所述，本研究开发了一种用于处理具有不同特征的时间序列的统计模型和实现该方法的软件。因此，我们认为本研究的大部分目的已经实现。然而，开发一个用户友好的实用软件还需要更多的研究。与本研究有关的论文和其他出版物已列入本研究项目的报告。还提供了开发软件的手册。

By the research in the previous fiscal years, we have developed a recursive filtering and smoothing algorithms for a general state space model which can be applied to diverse fields of time series modeling.Based on these results, in this year, we tested the developed program on various real data sets and made necessary modifications. We also did research on developing new models and more efficient algorithms. As a result of this research, based on the non-Gaussian state space model, a unified method fr analyzing time series with various characters such as the non-linearity, non-stationarity, non-Gaussianity, outliers and missing observations has been established. We also did research on the graphical display of the results of time series analysis and developed various programs for that purpose. In particular, by using the CALCOMP graphic libraries it becomes possible to use the graphical programs on various kinds of computers including mainframes and personal computers.As mentioned above, in this research a statistical model for handling time series with various character and a softrware for implementing the approach have been developed. Therefore, we believe that the most part of our purpose of this research has been achieved. However, more research is required to develop a user-friendly practical software.The papers and other publications produced concerning this research have been included in the report of this research project. A manual for the developed software is also available.