Selection of appropriate agreement statistic in binary diagnostic tests

Reliability plays an important role in the evaluation of agreement between observers. Therefore, Simple Percent Agreement Statistics, Cohen s Kappa, PABAK, Scott s agreement statistics, G-index and AC1 agreement statistics are used to calculate agreement between observers in binary diagnostic tests outcomes. This study aims to compare these agreement statistics and selection of the most suitable agreement statistics. For this purpose, MS Excel and IBMM SPSS Statistics 22 programs were used. The agreement statistics between observers were calculated in the tables where hypothetical data were generated according to the states of equilibrium and balance. It was observed that the best result was given by AC1 statistic when it is compared with the agreement statistics done on the same observer group after doing necessary arrangements according to the states of being affected by sensitivity, specificity and prevalence.