Ceiling effects in the Movement Assessment Battery for Children-2 (MABC-2) suggest that non-parametric scoring...

Initially designed to identify children's movement impairments in clinical settings, the Movement Assessment Battery for Children-2 (MABC-2) is also widely used to evaluate children's movement in research. Standardised scores on the test are calculated using parametric methods under the assumption of normally-distributed data. In a pilot study with thirty five 8-10 year old children (i.e., in Age Band 2 of the MABC-2), we found that maximal performance was often reached. These 'ceiling effects' created distributions of scores that may violate parametric assumptions. Tests of normality, skew, and goodness-of-fit revealed this violation, most clearly on three of the eight sub-tests. A strong deviation from normality was again observed in a sample of 161 children (8-10 years, Experiment 1), however ceiling effects were reduced by modifying the scoring methods, and administering items designed for older children when maximal performance was reached. Experiment 2 (n = 81, 7-10 years) further refined the administration and scoring methods, and again improved the distributions of scores. Despite reducing ceiling effects, scores remained non-parametrically distributed, justifying non-parametric analytic approaches. By randomly and repeatedly resampling from the raw data, we generated non-parametric reference distributions for assigning percentiles to each child's performance, and compared the results with the standardised scores. Distributions of scores obtained with both parametric and non-parametric methods were skewed, and the methods resulted in different rankings of the same data. Overall, we demonstrate that some MABC-2 item scores are not normally-distributed, and violate parametric assumptions. Changes in administering and scoring may partially address these issues. We propose that resampling or other non-parametric methods are required to create new reference distributions to which an individual child's performance can be referred. The modifications we propose are preliminary, but the implication is that a new standardisation is required to deal with the non-parametric data acquired with the MABC-2 performance test.