考慮治瘉者下多變量存活資料之迴歸分析

Abstract

家族或是多變量的存活資料常見於生物醫學的研究。此兩年計畫針對此類資料建立容許治癒或是免疫者之迴歸模型與統計方法。我們利用了混合模式將母體分為兩類子群體 (“會發病者＂與“免疫者＂)，以邏輯斯模式描述解釋變數如何影響發病與否的機率，以轉換模式描述發病時間與解釋變數的關係。我們並打算以二階段的方法處理推論問題。第一階段主要處理邊際模式參數的推論，重點放在邊際點估計量彼此相關性的估計上。第二階段加入了關連性結構，將討論發病與否的二元變量與發病時間兩種變數的關連性推估。

Multivariate survival data are commonly seen in biomedical applications in which each subject may experience multiple events of same or different types. In this two-year project, we consider regression analysis for multivariate survival data in presence of cure. The mixture framework will be adopted to formulate the marginal incidence and latency models. We will consider a two-stage procedure for statistical inference. In the first stage, we focus on marginal estimation which ignores the associations within the same sampling unit. The so-called sandwich method will be utilized to estimate the variance-covariance matrix for the marginal estimators. In the second stage, we will consider estimating the pairwise association parameters for both the susceptibility indicators and the latency variables within the same sampling unit.