2020 Fiscal Year Final Research Report
Semiparametric inferences for time-to-event data with incomplete data and their multidimensional extensions
| Project/Area Number |
17K00054
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Shiga University (2019-2020)
Kagoshima University (2017-2018) |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2017-04-01 – 2021-03-31
|
| Keywords | 生存分析 / 事象時間データ / 回帰分析 / 多変量分布 |
|---|
| Outline of Final Research Achievements |
We formulated the asymptotic distribution of correlated bivariate log-rank statistics and applied it to inference for bivariate event-time data with copula-type correlation. Some results were obtained in the studies of inference for the semi-competitive risk problem, and application of group-sequential bivariate log-rank statistics to sample size design. We also studied a semi-parametric estimation method using survival regression trees for relative survival, and proposed a Brier score for relative survival models to measure the predictive performance of regression trees, which was presented at a conference. We studied semi-parametric estimation of bivariate hazard models in which event-time and calendar-time are represented separately, and examined computational methods for the semi-parametric estimation.
|
| Free Research Field |
統計科学
|
| Academic Significance and Societal Importance of the Research Achievements |
医療の世界では，がんや循環器疾患などで治療効果を測るために，事象時間データの分析は必須である．経済分野では倒産などのイベントを分析したり，製造業分野では在庫がなくなるまでの時間を分析したり，事象時間データの分析の適用例は多くみられる．そのようなデータの分析方法について，Cox回帰モデルなどの生存時間データの統計解析法は必須であり，その当該分野において，現在まで得られている統計的な分析方法，理論，計算手法を発展させるための研究を行い，一定の成果を得ることができたことは，学術的意義をもつ．さらに，これらの手法を実際のデータに応用していくことで，社会的に還元をなすことができる．
|
