2021 Fiscal Year Final Research Report
Developing a theory for model selection in semiparametric statistical analysis
| Project/Area Number |
16K00050
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | The Institute of Statistical Mathematics (2018-2021)
Kyushu University (2016-2017) |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2016-04-01 – 2022-03-31
|
| Keywords | 因果推論 / 傾向スコア解析 / 情報量規準 / スパース推定 / セミパラメトリック推定 / 統計的漸近理論 / モデル選択 / SURE理論 |
|---|
| Outline of Final Research Achievements |
In causal inference, what is observed and what is actually observed usually influence each other, and applying classical statistical theory gives unreasonable results. One solution to this problem is to use propensity score analysis, which is rapidly developing, but the method of model selection, i.e., what regression model to use, has not been established. In this study, we developed an information criterion, a standard tool for model selection, for propensity score analysis and completed the general theory.
|
| Free Research Field |
数理統計学
|
| Academic Significance and Societal Importance of the Research Achievements |
因果推論の情報量規準としては，数理的に妥当でないものが試験的に用いられていたが，それは本成果の情報量規準と大幅に異なる値を返すものであった．つまり，両者のモデル選択の結果は相当に異なるものであり，本提案は今後標準的に用いられていくことが期待される．因果推論は機械学習・医学統計・計量経済学でのホットトピックであり，またモデル選択は統計解析において不可欠なタスクであるため，本成果の意義は小さくないものである．
|
