Lantz C A, Nebenzahl E
Life Chiropractic College West, San Lorenzo, California 94580, USA.
J Clin Epidemiol. 1996 Apr;49(4):431-4. doi: 10.1016/0895-4356(95)00571-4.
Two apparent paradoxes have been identified for the kappa (kappa) statistic: (1) high levels of observer agreement with low kappa values; (2) lack of predictability of changes in kappa with changing marginals. The first paradox is a function of prevalence of the trait in the sample, while the second is related to symmetry of observations in the disagreement categories. While examining the behavior of kappa as a function of the distribution of responses in a contingency table, it was discovered that for any measured level of observer agreement (Po) there are three characteristic values of kappa: kappa max, kappa min, and kappa nor, each of which is a function only of Po. The characteristic values allow an observed kappa (kappa o) to be placed into perspective. By observing symmetry in agreement and disagreement categories, the behavior of kappa is readily understood and predictable. We define symmetry expressions for agreement (SA) and disagreement (SD) in order to represent and quantify these effects. Kappa alone has little interpretive value and we recommend that studies reporting kappa also report Po, SD, and P++ (agreement on the presence of the trait).
已发现kappa（κ）统计量存在两个明显的悖论：（1）观察者一致性水平高但kappa值低；（2）随着边际变化，kappa变化缺乏可预测性。第一个悖论是样本中该特征患病率的函数，而第二个悖论与不一致类别中观察结果的对称性有关。在研究kappa作为列联表中反应分布函数的行为时，发现对于任何测量的观察者一致性水平（Po），都有kappa的三个特征值：kappa max、kappa min和kappa nor，每个特征值仅是Po的函数。这些特征值可以让观察到的kappa（kappa o）得到正确看待。通过观察一致和不一致类别的对称性，kappa的行为很容易理解和预测。我们定义了一致（SA）和不一致（SD）的对称表达式，以表示和量化这些效应。仅kappa本身几乎没有解释价值，我们建议报告kappa的研究也报告Po、SD和P++（关于该特征存在的一致性）。
