Estimating the Reliability of Students' Scores in University Courses on Tests during the Semester and the Extent of Influence of these Scores Based on Determining their Relative Weights

The aim of this study was to estimate the reliability coefficients of the tests prepared by faculty members in the preparatory year courses for both males and females at the University of Tabuk, using the Raju formula based on the predetermined fixed weights in the university regulations, regardless of their differentiating capability or the students' actual differences. Reliability coefficients were estimated again using the standard deviations of scores on these tests that were compared with the coefficients estimated by the Feldt-Gilmer formula that uses the variance covariance matrix of these tests. To address the questions of the study, the scores for all the students in 663 sections and 15 courses were analyzed, and the reliability coefficients were estimated. In light of the study’s hypothesis, another sample of 30 sections was selected randomly from the original sample to estimate reliability coefficients through the Feldt-Gilmer formula. The results indicated that the estimated mean of coefficients 0.68 in the average were lower than expected. Based on the university strategy which focuses on quality and excellence, higher reliability is required; mediocrity in this situation is not accepted. This means that 0.68 was compared to 0.67 and 0.69 estimated by the adjusted Raju formula using standard deviations as weights, and by the Feldt Gilmer respectively. The correlation between the reliability estimations using the last two formulas was very high (r = 0.98). The results revealed enough evidence statistically and practically for recommending the use of the modified Raju formula in practical situations to estimate the tests' reliability