Un modèle d'accord entre observateurs sur une échelle nominale
Résumé
By an appropriate hypothesis testing in a simulated example, one can show that Cohen's kappa can sometimes take values whose interpretation may be difficult. A new statistics mesuring interobservateur agreement is derived which does not involve pairwise considerations.This statistics is compared with kappa's measure and can be interpreted as a sample variance. Some properties relevant to Partition's Theory show that it values are ordered according to adequat agreement criteria.
Domaines
Applications [stat.AP]
Origine : Accord explicite pour ce dépôt