How serious is nonignorable missingness?
| Project/Area Number |
16K12402
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
Kano Yutaka 大阪大学, 大学院基礎工学研究科, 教授 (20201436)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
岩崎 学 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 (独)大学入試センター, 研究開発部, 教授 (10203829)
廣瀬 慧 九州大学, マス・フォア・インダストリ研究所, 准教授 (40609806)
菊地 賢一 東邦大学, 理学部, 教授 (50270426)
伊森 晋平 広島大学, 大学院理学研究科, 助教 (80747345)
|
|---|
| Research Collaborator |
MORIKAWA Kosuke 東京大学, 地震研究所, 特任研究員 (40824305)
IMADA Miyuki 日本電信電話(株), NTT未来ねっと研究所
TAKAGI Yoshiharu サノフィ(株), 統計解析・プログラミング部, 開発職
NAGASE Mario 大阪大学, 大学院基礎工学研究科
Kim Jae-Kwang Iowa State University, Department of Statistics, Professor
Yuan Ke-Hai University of Notre Dame, Department of Psychology, Professor
Jamshidian Mortaza(Mori) California State University, Department of Mathematics, Professor
|
|---|
| Project Period (FY) |
2016-04-01 – 2018-03-31
|
| Project Status |
Completed (Fiscal Year 2017)
|
| Budget Amount *help |
¥3,380,000 (Direct Cost: ¥2,600,000、Indirect Cost: ¥780,000)
Fiscal Year 2017: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2016: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
|
|---|
| Keywords | 無視不可能な欠測 / shared-parameter model / セミパラメトリック法 / NMAR / バイアス / 補助変数 / MLEのバイアス / 双方向因果モデル / モデルの識別性 / 脱落 / 対数正規モデル / 多重代入法 / 欠測メカニズム / NMARness / APB / 統計的因果推論 / 推定量のバイアス / MAR / 推定方程式の不偏性 |
|---|
| Outline of Final Research Achievements |
Under NMAR missingness, the observed likelihood, without a missing-data mechanism, leads to a biased MLE. In this research, we developed a new methodology to express the bias of the MLE due to the missingness in closed form. Using the formula, we provided several mathematical conditions under which inclusion of auxiliary variables reduces or inflates the bias.
The formula described above holds for any missing-data mechanism. This strong consequence can be proved because a shared-parameter model is taken for missingness.
A final contribution of this research to be reported is to take a semi-parametric way to relax the strong condition required for the conventional missing-data analysis.
|
Report
(3 results)
Research Products
(12 results)
