Personel alımında kullanılan bir yazılı sınav sonucunun genellenebilirlik kuramındaki farklı desenlerle karşılaştırılması
Özet
The purpose of this study is to conduct G and D studies by applying different designs in generalizability theory to the data obtained by evaluating the answers of individuals who participate in a job entrance exam consisting of open-ended items by different raters, and to analyze and interpret the G and Phi coefficients obtained as a result of these studies. The study group of the study was obtained with the group of 211 individuals participating in the open-ended items applied by ASYM with two raters in the Law subtest. In the test, they were asked to select to answer two of four questions. Individuals' answers were also evaluated by two raters. The data was limited by determining the items that individuals chose the most and a crossed design (bxmxp) was applied on the data created with 93 individuals. Individuals are cross-designed with items and raters and items are nested to raters in order to apply (bx(m:p)) the nested design, the data were limited and nested design was applied on the data obtained from 40 individuals. In order to analyze the sub-objectives of the study, the relative and absolute error variances and G and Phi coefficients were calculated by making G study with crossed and nested designs and D study was conducted. EduG program developed by Jean Cardinet in 2006 was used in order to estimate the variance components of the designs, calculate the variance explanation percentages of the components and conduct D studies. According to the results of the G study, the G and Phi coefficients calculated 0,33 and 0,29 for (bxmxp) design, and 0,76 and 0,64 for (bx(m:p)) design respectively. According to the results of the K study, it was observed that increasing the number of raters in the (bxmxp) design and increasing the number of items in the design (bx (m: p)) increased the reliability.