摘要
重复性、真实性和联系的测量应分别用组内相关系数(rI)、一致性相关系数(rC)和积矩相关系数(rP)。本文证明 rI≤rC≤rP,相等性仅在特殊情况下成立。在测量重复性、真实性和联系时,如果错用不恰当的相关系数,通常会高估重复性,低估联系,高估(如果用 rP)或低估(如果用 rI)真实性,本文相应地对几种 kappa 系数的应用亦作了讨论。
Repeatability,validity and association should he measured with intraclass correlation coefficient(r_I),concordance correla- tion coefficient(r_C)and product-moment correlation coefficient (r_P) respectively.It was demonstrated that r_I≤r_C≤r_P and the equality exists only in special condition in this paper.If inap- propriate correlation coefficients were used,when measuring repeatability,validity or association,repeatability would he usually overestimated,association underestimated and validity overestimated (if with r_p) or unedrestimated (if with r_I).The application of corresponding kappa coefficients have been also discussed in this paper.
出处
《中国卫生统计》
CSCD
北大核心
1995年第4期9-12,共4页
Chinese Journal of Health Statistics
关键词
相关系数
一致性
联系
重复性
真实性
医学统计
Correlation coefficient Agreement As- sociation Repeatability Validity Kappa