巴西生育資料之統計分析

Abstract

本篇論文主要的目的是利用人口與健康調查 (Demographic and Health Surveys) 計畫蒐集的巴西嬰孩生育與死亡的資料進行統計分析。巴西的領土與人口為世界前五大的國家，資料存在相當程度的異質性。我們將資料分為六個區域進行分析與比較。首先我們採用了倖存分析與考慮治癒模式的倖存分析，探討影響巴西嬰孩存活時間的風險因子，亦利用脆弱性模式處理資料存在的群聚關聯性。我們亦採用了羅吉斯迴歸模式，探討巴西嬰孩的前三年死亡率，也以脆弱性模式處理資料存在的群聚關聯性。最後我們利用無母數方法與廣義線性混合模式，探究母親年齡與子女個數的關係。

In the thesis, we apply different statistical methods to analyze Brazil’s birth data, a survey conducted by Demographic and Health Surveys (DHS) program. Brazil is the fifth-largest country by both area and population in the world. To adapt to the heterogeneity, we conduct separate analysis based on six regions and then compare the results. First, we apply survival analysis and cure mixture model to investigate possible risk factors of infant’s mortality. The frailty approach is adopted to account for the cluster effect. Then we apply logistic regression analysis to study the probability of death by the first three years. Finally, we investigate the association between mother’s age and the total number of children based on nonparametric regression analysis and Poisson regression models.