论文部分内容阅读
Gross(1986)提出了一种修改型的Kappa系数以作为当评判者数相对于受评者数和分类数较大时总一致性的测度。出于同样的目的,本文也相应地提出了一种新的Kappa型测度K′_F。该Kappa被证明与Light和Margolin(1971)的连带性测度R~2是等价的,从而Light和Margolin(1971)的C统计量也可作为机遇一致性零假设的检验统计量。本文还证明了K′_P在备择假设下的大样本分布是近似于正态分布的,并给出了K′_F的近似均值和方差;与此同时,本文也对Gross的Kappa在备择假设下的近似方差计算公式作了更正。至于K′_F与Gross的Kappa之间的关系和比较本文也简单地作了说明。
Gross (1986) proposed a modified Kappa coefficient as a measure of the overall consistency when the number of judgers is large relative to the number of reviewers and the number of categories. For the same purpose, this paper also proposes a new Kappa-type measure K’_F. The Kappa proved to be equivalent to Light and Margolin’s (1971) coherence measure R~2, so Light and Margolin’s (1971) C statistic can also be used as the test statistic for the opportunity-consistent null hypothesis. This paper also proves that the large sample distribution of K′_P under the alternative hypothesis is similar to the normal distribution, and gives the approximate mean and variance of K′_F. At the same time, this article is also for the Kappa of Gross. Assume that the approximate formula for calculating the variance has been corrected. As for the relationship between K’_F and the Kappa of Gross and comparisons, this article also briefly explains.