2018-08-012024-05-16https://scholars.lib.ntu.edu.tw/handle/123456789/669071摘要：針對遞迴資料發生率函數，我們將精準刻劃解釋變數之空間維度及不做假設於遞迴 事件之相關結構，在所考慮之降維&#24315;歸模型下，此計劃第一個研究主題乃提出新的 統計推論，除此之外，估計方法之執行亦欲發展出有效能的計算流程。此研究的第 二主題乃針對發生率函數之結構參數提出完整的檢定方法並驗證檢定統計量之模型 選擇一致性，預期所提之估計方法將有助於模型結構檢測方法之發展。最後，我們 將做一系列數值模擬於估計及檢定之有限樣本性質之探究並應用於實際遞迴資料研 究上。<br> Abstract: A dimension reduction regression model is proposed for the conditional mean of a recurrent event process on covariates of interest over time. In this research issue, the dependence structure is allowed to be arbitrary among recurrent events and the covariate space is precisely characterized by the central mean subspace. A dimension reduction approach is expected to be developed to simultaneously estimate the structural mean dimension, central mean subspace, and occurrence rate function. Further, such an estimation should be effectively carried out by a forward algorithm with respect to the dimension. Without assuming a particular parametric or semiparametric model formulation on the occurrence rate function, we also plan to build test rules for the rate, shape, and size parameters. The finite-sample performance of the proposed estimators and testing procedures will be assessed through comprehensive simulations and some empirical examples.中央平均子空間模型選擇一致性遞迴性資料發生率函數.central mean subspacemodel selection consistencyrecurrent event processoccurrence rate function