家庭收支調查抽樣設計簡介

為使樣本之特徵結構更接近母體，提高樣本代表性，以及考量經費及人力負擔，本調查抽樣方法採分層兩段隨機抽樣，先對全部村里分層，並以村里為第一段抽樣單位，樣本村里中之戶為第二段抽樣單位。分層方法原以各村里就業人口之產業結構為分層變數，採特定比例按縣市將村里分成都市、城鎮及鄉村三層。然隨著社會環境變遷，分層方式需隨之修正，改以相關分析及主成分分析選取出年齡、教育程度及就業人口之產業結構等變數，以集群分析對村里重新分層，藉以提升分層效益。另考量調查環境困難，為使調查作業順利執行，採重複的橫斷面抽樣設計，即每年均以相同的抽樣方法重新抽樣並進行調查。

關鍵字

分層兩段隨機抽樣；重複的橫斷面抽樣設計；相關分析；主成分分析；集群分析

並列摘要

In order to let the structure of sample approximate population, increase representative sample, and consider economic and manpower, the stratified two-stage sampling method is adopted by survey of family income and expenditure. First, all of TSUN/LIs are stratified. The TSUN/LIs are the primary sampling unit (PSU), and the household in the PSU is the secondary sampling unit (SSU). Formerly, the method of stratification depends on some criteria to classify TSUN/LIs into village, city, as well as town stratum. However, due to social environment transition, the method of stratification needs to be revised. In the first place, we select age, education level, structure of employed population as the stratified variables by using correlation analysis and principal components analysis. Then we use the above variable and cluster analysis to let TSUN/LIs classified into stratum in order to promote efficiency. Considering the environment of statistical survey, we use repeated cross-sectional design to carry out the investigation. The repeated cross-sectional design means that we use the same sampling method to generate annual sample and conduct the survey.

並列關鍵字

stratified two-stage sampling ； repeated cross-sectional design ； cluster analysis ； correlation analysis ； principal components analysis ； cluster analysis