부산대학교 도서관

Two series of measurements taken by one investigator and by several investigators on the same subjects were compared to examine the reliability of the data obtained in large-scale somatometric research. The series of differences was analysed using analysis of variance, and statistically significant inter-observer difference was observed in 17 out of 20 variables (Table 3). From a practical point of view, 7 out of 17 variables with statistically significant inter-observer difference should be excluded from analysis because of their great error variance (Table 4). It was suggested that the direct measurements were more precise than the subtracted measurements of the upper limb. The examination of possible effects of measurement error on several biological distances suggested that shape as well as size distance is affected by systematic measurement error. The influence of measurement error can be minimized by excluding the measurement items having great measurement error.