台灣六都之健保就醫支出地區性差異:以變異數成分模型分析

Abstract

一般線性模型是統計學上在對資料進行變異數分析時常用的模型，以往通常將欲估計的參數以固定效應的形式進行估計；而變異成分模型是階層線性模型的一種。在生物統計學中，我們以固定效應和隨機效應分別代表不同參數的效應，再對特定因子進行變異數分析。在本文中，我們先回顧兩種模型的估計方式，再對台灣六都西元2012年健保資料庫的醫療費用與個人資料之間的關係進行分析，利用兩種模型比較不同直轄市與行政區於醫療費用的差異，再對變異成分模型進行兩種得分檢定，了解地區差異。最後再從實際數據分析的結果進行說明。In statistics, general linear model is a common model to execute analysis of variance; in the past, we usually estimate the parameters in the form of fixed effects. The variance component model is also named as the random effects model, and is a kind of hierarchical linear models.We use “fixed effects” and “random effects” to denote the effects of different parameters, and then making inference on these effects. In this thesis, we first review the ways to estimate the two models, and then analyze the difference between medical treatment expenses in six municipality of Taiwan in 2012 adjusting for basic personal information of patients And then use two kinds of score test to test the difference among municipalities. Finally, we provide a conclusion for the data analysis.