We find that in the second case, it has a greater similarity between A and B than in the first. Indeed, although the percentage match is the same, the percentage match that would occur “randomly” is significantly higher in the first case (0.54 compared to 0.46). In most imaging techniques, each patient can bring several positive results, and the data is of course aggregated within patients. Clustering has no influence on the calculation of the free-response kappa, but must be taken into account when calculating the standard error. It is important to realize that the global kappa with free response is a weighted average of kappa statistics within the cluster, with weights proportional to b k + c k + 2d k, the total number of positive scores in a cluster (without linkage). This decomposition applies to each partition of the data and can be performed for any covariate, e.B. comparing agreement beyond chance in obese and non-obese patients or skeletal lesions versus soft tissue lesions. The following table shows the chances of a random match between the correct answer and the evaluator`s score: The free-response kappa provides an estimate of the match beyond chance in situations where only positive results are reported by the evaluators. In an attribute agreement analysis, the kappa measure only takes into account the random agreement.
Nevertheless, significant guidelines have appeared in the literature. Perhaps the first Landis and Koch, who characterized the values < 0 as no match and 0–0.20 as weak, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1 as almost perfect agreement. .