摘要
本文以12个非省会市县的800个城镇居民家庭消费支出中的8个指标的调查数据为例,验证了多目标分层等距复合抽样方法。第一阶段采用Q-型系统聚类方法进行分层,用欧氏距离平方法计算样本距离,用类间平均链锁法计算小类之间的距离。结果表明在样本容量相同的情况下,多目标分层抽样法抽取的样本均值与总体均值间的误差明显小于多目标简单随机抽样。第二阶段是在聚类分层的基础上采用随机起点对称等距抽样。结果表明在相同样本条件下进行多目标抽样时,当随机抽样相对误查很大时,分层等距抽样复合抽样比分层抽样更具优势,能够大幅降低抽样误差。
Based on the survey data of 8 indicators of 800 urban households consumption expenditures in 12 cities and counties which are not provincial capitals, multi-objective stratified systematic compound sampling method is verified. In the first stage, Q-Hi- erarchical cluster analysis method is adopted; the sample distance is calculated with the Squared Euclidean distance, and the minority class distance is calculated with the Between-groups linkage method. The results show that under the condition of the same sample size, the errors between sample mean and population mean of multi-objective stratified sampling method are smaller than that of ran- dom sampling. In the second stage, random start symmetric systematic sampling is adopted based on the hierarchical clustering. The results show that under the same sample conditions, when sampling for multi-objective, stratified systematic compound sampling is better than random sampling, and can decrease the sampling errors significantly.
出处
《西部金融》
2013年第9期34-38,共5页
West China Finance
关键词
简单随机抽样
多目标分层等距复合抽样
样本均方差
simple random sampling
multi-objective stratified isometric composite sampling
sample mean square error
