To present statistical methods on appropriate data analysis from complex surveys and errors arising from ignorance of weights or design of samples. We took Chinese National Nutrition and Health Survey in 2002 as an example to analyze the prevalence of hypertension among population aged 15 and over. We used four combinations of analyses, including with or without weighting or considering sample designs. If weights is omitted, it would result in biased prevalence estimates and also influence thern estimates of standard errors. While omitting sample designs would result in underestimating the standard error estimates and then testing the false positive hypothesis. Through appropriate analysis, we found Chinese people in large-sized cities had the highest prevalence of hypertension (28.77%, 95% CI:rn 25.69%-31.84% ) while people in the poorest rural area having the lowest prevalence of hypertension ( 14.21%, 95 % CI : 12.64%- 15.79% ). The prevalence of hypertension among people in small and medium-sized cities and other rural areas ranged from 20.48 % to 24.37 % with statistically insignificant difference. It is necessary to use appropriate methods to analyze data from complex surveys.%提出复杂抽样调查数据的分析思路和方法以及忽视权重和抽样设计时会出现的问题.文中以2002年中国居民营养与健康状况调查数据中高血压患病率的估算为例,分加权和不加权、考虑和不考虑整群设计特征的四种组合情况对数据进行分析.表明忽视权重的设置会影响点估计和标准误的估计,忽视对整群设计特征的考虑不仅会高估结果的精确度,还会得到地区间患病率有差异的假阳性结论.因此使用合理的统计方法分析复杂抽样调查数据非常有必要.
展开▼
