Comparison of Interesting Reliability of Damage Social Behavior scale according to classical testing theory and generalization theory

In this study, it was aimed to compare the reliability of the raters between the classical test theory and generalizability theory and determine which theory provides more information. The study group consisted of twenty children with autism who had been studying in a special education and rehabilitation center in Sakarya for at least one year. In the study, "damaging social behaviors"subscale of the Autism Social Skills Profile scale. The two raters who took part in the study scored twenty individuals according to twelve items on the scale. The data obtained were calculated by using the internal consistency reliability of the raters in classical test theory. Within the scope of generalizability theory; The raters have been designed together to design the individuals that they score together and alternately. One of the designs designed in this context is the bxmxp cross pattern in which the individuals (b) are scored by each of the raters (p) in the direction of the same items (m). The other is that the individuals are scored by different raters in line with all items (p: b) xm nested pattern G and K studies were conducted separately. At the end of the research, it was found that the Cronbach Alpha coefficients, which indicate the internal consistency reliability of the raters according to the Classical Test Theory, were acceptable (first and second raters were 0.70 and 0.68, respectively). The correlation coefficients between the scoring and Kendall's coefficient of coherence were found to be high. The G and Phi coefficients, calculated separately for the cross and nested pattern within the scope of generalizability theory, are acceptable. (For the bxmxp pattern G = 0.69 Phi = 0.64; (p:b)xm pattern G = 0.73 and Phi = 0.69). It was concluded that increasing the number of items was more effective than increasing the number of raters. According to all these results, it was determined that generalizability theory gave more detailed information than classical test theory, and the level of consistency among the raters was high in both theories. In studies where the source of variability is the only one, it is recommended to use the General Test Theory and the Generalizability Theory in studies where more than one source of variability is available. Keywords:Autism, Damaging social behavior, Classical test theory, generalizability theory, interrater reliability