2022 Fiscal Year Final Research Report
Developing a robust method for inferring changes, relations and group differences in longitudinal data: Applications to psychological research
| Project/Area Number |
19K14378
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 10020:Educational psychology-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Satoshi Usami 東京大学, 大学院教育学研究科(教育学部), 准教授 (20735394)
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Keywords | 縦断データ / 個人内関係 / 交差遅延パネルモデル / 周辺構造モデル / 構造ネストモデル / 測定誤差 / 相互関係 / 構造方程式モデリング |
|---|
| Outline of Final Research Achievements |
The outline of research achievements is as follows:
(1) To infer the within-person relation in longitudinal design the applications of RI-CLPM are rapidly growing. Using the causal inference framework proposed in epidemiology, a new estimation method has been developed that is robust to the model misspecifications, including violation of linearity between focal variables and time-varying observed confounders. The paper was accepted in major international journal. (2) A paper that discusses some methodological issues in applying another statistical model (GCLM) for inferring within-person relation was accepted in major international journal. (3) An overview article that summarized various statistical models and their relations used for inferring within-person relation was accepted in a domestic journal. (4) An overview article was submitted that explains and illustrates the method proposed in (1), also we extended it to the case where observations are influenced by measurement errors.
|
| Free Research Field |
心理統計学
|
| Academic Significance and Societal Importance of the Research Achievements |
縦断データを用いた応用研究は年間で一万件以上世界で報告されており、中でも縦断的に測定された変数間の関係性の推測は主要なテーマである。本研究では、異なる研究領域や文脈を通してこれまで提案されてきた様々な統計モデル間の概念的・数理的関係性を整理して、特に個人内関係の推測上生じうる問題点や他のモデル設定の可能性を示した。また、交絡変数に対する線形性の仮定など統計モデル上生じうる誤設定に対して頑健かつ柔軟な個人内関係の推測方法を提案し、大規模縦断データへの適用例とともに、一定の時点数の下で十分な推定性能を有することを示した。
|
