回溯性生育資料統計分析

Abstract

本論文的研究動機來自一筆孟加拉的生育資料 (Bangladesh Demographic and Health Survey, 簡稱 BDHS)。為研究生育的間隔時間，由至少生育兩胎的婦女取樣。我們對這筆資料進行基本分析時，發現有些資訊可能帶有偏誤，我們將之歸類為右截切的問題。我們進行了模擬實驗，探究截切所造成的影響，並透過反轉時間軸的方式成功調整右截切的偏誤。最後我們將此方法應用在資料分析，先前的偏誤已不復見，亦看出婦女生育的間隔時間的世代差異。

This thesis was motivated by a real dataset published in Bangladesh Demographic and Health Survey (BDHS) about women’s birth and other health information in Bangladesh. To study serial birth intervals, we only selected women who gave at least two births which however created bias in the sampling procedure. We formulate the problem under the framework of right truncation and analyze its effect via simulations. By applying the reverse-time technique, we can correct the truncation bias. The modified method is then applied to analyze the data. The new result shows that younger and older generations do reveal some difference in their first birth intervals.