2015 Fiscal Year Final Research Report
New developments of missing data analysis: NMARness and APB
| Project/Area Number |
25540011
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
Yutaka Kano 大阪大学, 基礎工学研究科, 教授 (20201436)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
IWASAKI Manabu 成蹊大学, 理工学部, 教授 (40255948)
|
|---|
| Co-Investigator(Renkei-kenkyūsha) |
TAKAI Keiji 関西大学, 商学部, 准教授 (20572019)
OTSU Tatsuo 大学入試センター, 研究開発部, 教授 (10203829)
HIROSE Kei 大阪大学, 基礎工学研究科, 助教 (40609806)
KAMATANI Kengo 大阪大学, 基礎工学研究科, 講師 (00569767)
KIKUCHI Kenichi 東邦大学, 理学部, 教授 (50270426)
|
|---|
| Research Collaborator |
Sobel Michael E. Columbia University, Professor
Yuan Ke-Hai University of Notre Dame, Professor
Ricardo Silva University College London, Lecturer
Mortaza Jamshidian California State University, Fullerton, Professor
Aapo Hyvarinen University of Helsinki, Professor
|
|---|
| Project Period (FY) |
2013-04-01 – 2016-03-31
|
| Keywords | Missing at random / APB / NMARness / bias of the MLE / sarrogate endpoint |
|---|
| Outline of Final Research Achievements |
This research project has been completed by the two research groups conducted by Professor Yutaka Kano and Professor Manabu Iwasaki. We have offered research colloquiums several times for each year to advance the research project. The aim of the research project is to re-structure the theory of missing data analysis and to apply them to some statistical models for the analysis with missing data. Results of the project include mathematically weakening the MAR condition, defining NMARness and Approximate population Bias (APB) and studying mathematical properties of the NMARness and APB. Applying these theoretical results, we studied effectiveness of introducing auxiliary variables in several statistical models for the analysis of missing data. One particular result is to derive mathematical conditions under which introducing surrogate endpoints can reduce the bias of the MLE for data with possibly missing data at the endpoint.
|
| Free Research Field |
統計科学
|
