Analysis of the Multiraters Agreement with Log-Linear Models

OpenAlex · İngilizce

In this study, 87 digital panoramic images are classified by the three raters to assess the accuracy of diagnosis of peri-implant bone defects. The coefficient of kappa is obtained as 0.81 among the three raters, which indicates an almost perfect agreement. Then, the log-linear agreement models are applied to the data. The best model is determined based on the model selection criteria. Using the best model, we estimate the agreement parameter. It is 33 times higher for three raters to make the same decision than to make a different decision. The results show that the coefficients of the agreement only show the value of the fit between raters. On the other hand, agreement models provide a model equation for the raters, and more detailed and consistent results can be obtained by calculating the agreement and association parameters.