Psychometric properties of child- and teacher-reported curl-up scores in children ages 10-12 years

The purpose of this study was to examine the psychometric properties of chi ld- and teacher-reported curl-up (CU) scores in children ages 1012 years in both a norm-referenced (NR) and criterion-reference (CR) framework. Eight- Four children, 36 boys and 48 girls, performed the FITNESSGRAM (Cooper Inst itute for Aerobics Research, 1992) CU test on 2 days separated by 48-72 hr. Two video cameras were used to record students' CU performances. Two stude nts performed the CU at the same time, with each child performance recorded by one camera. The test was terminated when the child stopped due to fatig ue or after two form errors occurred. Teacher-reported scores were the aver age of two independent ratings of each video performance, while child-repor ted scores came from data collected and recorded by the children. Single tr ial norm-referenced reliability was R = .75 for girls and R =. 80 for boys for teacher-reported CU and R = .69 and R = .70 for child-reported CU for g irls and boys, respectively. CR reliability was examined using P, proportio n of students who consistently passed or failed the test across 2 days, and k(m), defined as reliability with chance removed. For teacher-reported sco res, P = .89 and k(m) = .58 for boys and P = .81 and k(m) = .62 for girls. For child-reported scores, P = .86 and k(m) = .72 for boys, while P = .79 a nd k(m) = .58 for girls. For teacher-reported data, 39% of boys passed and 50% failed the test on both days, while for girls the percentages were 27% pass and 54% fail. For child-reported data, 64% of boys passed and 22% fail ed on both days, while 54% of girls passed and 25% failed. NR validity was examined by correlating teacher and child-reported scores. The resultant co efficient was r = .42 (95% CI = . 11-.66) for boys and r =.67 (95 % CI =.58 -.74) for girls. Additionally, child-reported scores were significantly hig her than teacher-reported scores. CR validity was examined with a contingen cy coefficient, and results indicated C = .55 with 44% false master errors for boys and C = .65 with 29% false master errors for girls. The findings o f this study suggest that while NR reliability estimates were moderate for teacher-reported scores, single trial estimates suggest that child-reported CU should be viewed with caution. In regard to CR reliability, both teache r-reported and child-reported reliability were moderate. However there were marked differences between teacher- and child-reported scores, with childr en reporting higher percentages of students passing and lower percentage of student failing the test when compared with scores reported by teachers. V alidity was rather moderate when viewed in either a NR and CR framework. It is suggested that problems with child-reported scores may De due to the ne ed for additional practice or simplification of the testing protocol.