2021 Fiscal Year Final Research Report
Statistical inference for random-effects meta-analyses under model misspecification
Project/Area Number |
19K20229
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 60030:Statistical science-related
|
Research Institution | Keio University (2021) The Institute of Statistical Mathematics (2019-2020) |
Principal Investigator |
Nagashima Kengo 慶應義塾大学, 医学部(信濃町), 特任准教授 (20510712)
|
Project Period (FY) |
2019-04-01 – 2022-03-31
|
Keywords | メタアナリシス / 変量効果モデル / 予測区間 / モデル誤特定 |
Outline of Final Research Achievements |
We updated the R package, pimeta, which was previously released to calculate prediction intervals for meta-analysis. The improved pimeta package can also apply prediction intervals not applicable in other software. We released a preprint of a software paper describing estimation methods applicable in the pimeta package and how to use the pimeta package. In addition, we investigated the statistical properties of prediction intervals under model misspecification in the random-effects model. As a result, the bootstrap prediction interval (Nagashima et al., 2019) can control coverage rates better than other methods under various conditions. Moreover, we showed that Wang & Lee's (2019) non-parametric prediction interval is entirely different from other methods (i.e., Higgins et al. 2009, Partlett & Riley 2017, and Nagashima et al. 2019).
|
Free Research Field |
生物統計学
|
Academic Significance and Societal Importance of the Research Achievements |
メタアナリシスの適用結果は,医療政策や診療ガイドラインの策定などの公的な用途にも活用されており,社会に与える影響は非常に大きいと考えられる.本研究の成果により,既存の方法論よりも正確なエビデンスを提供することが可能となるため,学術的に非常に大きな貢献が期待できる. また,メタアナリシスで用いる変量効果モデルは,線形混合効果モデルの一種である.線形混合効果モデルは,幅広い領域で用いられる広範なモデルを含んでいる.モデル誤特定下での予測区間への影響は今までに検討されていない.本研究の成果により,広いクラスで適用可能な新しい理論的知見が得られると期待できる.
|