2023 Fiscal Year Final Research Report
Inferences and machine learning methods for multivariate time-to-event data with incomplete information
| Project/Area Number |
21K11783
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 60030:Statistical science-related
|
|---|
| Research Institution | Shiga University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2021-04-01 – 2024-03-31
|
| Keywords | 生存時間解析 / コピュラモデル / 計数過程 / ランダム効果解析 / 層別解析 / 木構造モデリング / 多変量標本分布 / 計算機統計 |
|---|
| Outline of Final Research Achievements |
(1) We studied the refinement of the inference theory of the normal random effects model of between-trial variation in stratified analysis, and applied our research on exact computation by computational algebra statistics to the log-rank test in the Mantel-Haenszel type stratified analysis method for binary data. (2) We studied multiple assignment methods for the use of time-varying covariates in Cox regression, and extended our research on group sequential design for bivariate survival time models. For bivariate event-time data with copula-type correlation structure, we studied the theory and methods of inference for semi-competitive risk problems, and conducted research on computational theory for NPMLE and modification of log-rank statistics. Some solutions to various problems in these studies were given and summarized in research papers and other publications.
|
| Free Research Field |
医学統計，数理統計，機械学習
|
| Academic Significance and Societal Importance of the Research Achievements |
層別解析は交絡調整の伝統的な統計手法だがランダム効果モデルといった多用される方法論においても，その標本分布は精緻化されていない側面があったが，本研究によりそれらの問題を有意義に解消する基盤が創出できた．多次元データのとり扱いに必要となる多変量分布論において多変量正規分布を有意義に超えるものは多くないが，本研究ではコピュラ型相関をもつ2変量分布をイベント時間データに定式化し，医学統計の応用において興味ある展開のいくつかを惹き出した．今後において決定木やランダムフォレストといった機械学習の方法に層別解析の方法論を融合させていくための足掛かりを得ることができ，学術的および社会的意義をもつ．
|
