摘要
目的：医学研究中经常遇到重复观测数据的统计分析问题，由于重复观测数据间存在一定的自相关性和随机误差的多层性，因而其分析方法有别于一般的统计分析方法。方法：对重复观测数据协方差矩阵的结构特征进行检验，对统计模型的选择、生长曲线转化和正交对比变换等方法在参数估计、算法实现和结果解释等方面进行了综合比较。结果：提出了重复观测数据统计分析基本策略，完成了用ＰＡＳＣＡＬ语言编制的专用软件ＲＥＰ，并取得了一批医学应用成果。ＲＥＰ软件有七个功能菜单。结论：本研究成果值得在医学研究中广泛推广和应用。
Objective:Repeated measures data often arise in medical research.Owing to correlation among repeated measures and its hierarchical error,the statistical analysis with repeated measures is not quite the same as the classical approach.Methods:The structure types of covariance matrices of repeated measures data were tested.The approaches such as the selection of statistical models,growth curve transformations and orthogonal contrast transformations were compared comprehensively in the estimation of parameters,the implementation of algorithms and the explanation of statistical results.Result:A basic analytic strategy was proposed and a proper software named REP was implemented in PASCAL language,REP has seven menus in it .A set of outcomes with application to medical research were obtained.Conclusion:The results in this study deserve popularizing and applying widely in medical research.
出处
《中国卫生统计》
CSCD
北大核心
1998年第6期5-8,共4页
Chinese Journal of Health Statistics
基金
国家自然科学基金
