在實際的應用研究中，部分變量的數據可能不易取得或花費昂貴，因而造成研究工作的重大瓶頸。爲了使研究易於進行，二階段抽樣的研究設計便應運而生。在二階段抽樣設計下，研究者先於研究之第一階段對所有樣本個體取得易於測量之變數數據，再於研究之第二階段對部分個體取得不易測量變數之數據。因此，在二階段抽樣設計下，研究者所取得之數據型態可視爲是「缺失數據」或「測量誤差數據」的特例。本文僅就文獻中針對二階段抽樣設計之迴歸分析方法學進行簡要的回顧，同時展望二階段抽樣設計相關統計研究課題的未來發展方向。 In many fields of applications, data on some variables under study may be difficult or expensive to collect, causing an important bottleneck in research work. To bypass the difficulty, the two-stage sampling design has proved to be a useful strategy. With this sampling design, at stage Ⅰ of the study data on some variables that are easy to measure are collected for all study subjects, while full and exact data are obtained only for a selected subset of the whole sample at stage Ⅱ. Consequently, data collected under a two-stage study are "incomplete" in the sense that the data are incompletely observed for all subjects, except for members selected at stage Ⅱ. The set-up can thus be viewed as a special type of "missing" or "measurement error" data. Regression analysis under two stage sampling design has been an active research area in statistical literature for two decades. This paper briefly reviews the existing methodologies for this problem, and indicates potential future research directions in this area.