مرکز منطقه ای اطلاع رساني علوم و فناوري - THE BAYESIAN APPROACH TO FIND OUT THE AGREMENT FOR MULTIPLE RATERS AND MULTIPLE CATEGORIES OF SPORTS ACTIVITIES

Abstract :

If the N subjects are assigned independently to one of k categories by multiple separate judges or raters inter and intra observer variability exits. There are two sources of variability. The first occurs when the observers identify and localize the object of interest, and the second happens when the observers don’t make appropriate measurement on the object of interest because in many circumstances, the categories into which subjects are classified do not have precise objective definitions. Therefore the assessment of agreement among raters is very important. A number of statistical methods are available to quantify the degree of agreement between measurements made by different observers. When the measurements are categorical the associations among categorical variables can be obtained by some classical approaches. The Bayesian approach will also be employed to provide statistical inferences based on various models of intra and inter rater agreement. The difference of the conventional approach and the Bayesian approach, the elements of a Bayesian application are explained and prior information, experimental information, the likelihood function, the posterior distribution, and the predictive distribution is included. Most statistical software packages don’t cover Bayesians statistical approaches, therefore the software package WinBUGS can be used to implement Bayesian inferences of estimation and testing hypotheses. This work aims to use the Bayesian approach to calculate and interpret the degree of agreement of multiple raters for the assessment of sports-related activities. If there is high agreement between raters, just why and how they indeed agree? On the other hand, when they do not agree, it is important to know why. An analysis of agreement should always be followed by a test for homogeneity between raters. We focused also on disagreement between raters. Therefore the pairwise agreements were also inspected. When multiple raters and multiple outcomes are encountered the Bayesian approach was is a good statistical technique to assess the agreement among raters.