Study on the creation of a new ignobility condition and the investigation of an estimator distribution in the analysis of missing data

2017 Fiscal Year Final Research Report

• Project/Area Number: 26730022
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: Kansai University
• Principal Investigator: TAKAI Keiji 関西大学, 商学部, 准教授 (20572019)
• Project Period (FY): 2014-04-01 – 2018-03-31
• Keywords: 欠測データ / MAR / 条件付き独立 / EMアルゴリズム / 判別分析 / 漸近理論

Outline of Final Research Achievements

In this project, I developed the theory for analysis of missing data and applied it to special data in the discriminant analysis. As theoretical study research, I derived conditional independences equivalent to MAR(missing at random) under monotonic and non-monotonic missing-data mechanisms. In addition, I constructed a method to overcome some difficulties in computation and estimation of the parameters of interest with missing data by using the selection matrix. It allows us to investigate properties of the estimator which are necessary for inference. As application research, I tackled a semi-supervised learning problem in discriminant analysis using the missing-data analysis theory. The semi-supervised learning is an estimation of the parameters from the partially observed data. I showed that the use of the missing-data analysis theory makes it possible to obtain the correct discriminant rule even with the such partially observed data.

Free Research Field

統計科学