The purpose of this study which is follow up study of Lee and Kim(2015b)'s was to analyse error sources and estimation of reliability in peer review of forced connection method-sportscasting by applying generalizability theory. Generalizability theory quantify error sources of the data measured under certain specific situation set by the researchers. It is an analysis method that the relative influences of each error sources taking from score is determined(G-study), and the effective measurement condition future applicable is provided(D-study). Participants were 10th high school students(N=216). Data were collected from student's peer review results and analyzed using univariate and multivariate generalizability theory. Results showed that error source for video have a more significant impact than other error sources. But the result by analyzing the gender difference was that error source for the interaction of video and participants have a more significant impact than other error source in the case of girls. Peer review used in this study showed high generalizability coefficient and even when reducing the number of video or participants it can maintain the adequate reliability. But generalizability coefficient of boys was higher than girls and specific measurement conditions leading to enhanced reliability were different when analyzing by gender difference. Also, method of analysis which cannot reflect measurement conditions properly estimates the reliability excessive. Discussions were provided in term of the relative influences of each error sources, the effective measurement condition maintaining the Generalizability coefficient of a certain level, and the comparison the Generalizability coefficient with the way of estimation traditional reliability applying univariate and multivariate Generalizability theory taking from score in peer review of forced connection method-sportscasting.
The purpose of this study was to analyze and confirm whether the items used in final paper and pencil test was determined to DIF when school sports clubs in each school operated by discriminatory curriculum in accordance with gender. Participants were 8th middle school students(male=135, female=141). They joined in school sports club every week from freshman to sophomore 1st semester. At that time, boys of them participated in soccer and basketball, and girls played dodge ball. They studied soccer unit at sophomore 1st semester, and had a final examination consisting of 5 soccer items. Using the data, differentially functioning item by the population difference between male and female were analysed quantitatively and qualitatively. The results showed that Mantel-Haenszel method(using classical test theory), comparison of item characteristic curve and likelihood ratio test(using IRT) determined item number 4 and 5 to differentially functioning item. Finally, item number 4 were identified differentially functioning item in favor of male students in intensive qualitative analyses. That item have low content validity and application-level of cognitive behavior classification. The result provides that application-level item can be functioning differentially to female students with little sports experience than male students in paper and pencil test of PE.
