|Budget Amount *help
¥4,100,000 (Direct Cost: ¥4,100,000)
Fiscal Year 1989: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 1988: ¥3,300,000 (Direct Cost: ¥3,300,000)
The purpose of the present project is to study theories of a knowledge-system for statistical data analysis and to construct concretely the software system, to be used commonly on personal computers.
As a whole, the project progressed favorably in research and development. Firstly, an intelligent support system was completed successfully for a sampling survey with very complicated structure, as shown in complex questionnaires, and was presented at domestic and international meetings. Secondly, for an intelligent system of numerical analysis. methods or maximum likelihood estimation are considered, and a system MLE-SYS was studied and constructed on a personal computer, like NEC PC-9801, J- 3100 and so on. The MLE-SYS system was able to have the most useful 22 analytical methods 'With intelligent know-how to apply their numerical methods to data. this system was also presented with several illustrations at an international meeting. Thirdly, regarding statistical inference based or) sever
al kinds of data, practical determination of processes of statistical methods was investigated theoretically and confirmatively in view of the optimality. the inference procedures of statistical methods are essentially based on phenomena, and how to recognize the mathematical model is our goal.
Therefore we made our most effort for this inference problem in view of statistical properties of the process. Thus the software system of intelligent system-lead type was studied, and was immediately realized to Micro-NISAN system we have developed for several years. As the application) of this system, many newly-developed sampling plans with inspection-process were proposed. All of these results were presented partially at intenational meetings, where some were invited. The Micro-NISAN system is avilable to educate mathematical statistics, too. Fourthly, analytical methods were studied for latent binary trait class, and were obtained to solve the problem. Fifthly, symbolic algorithms were studied for multivariate analysis, and the obtained results were presented as an invited topic at an international meeting. (See more details in the respective papers shown in this Report.) Less