The applicability of standard error of measurement and minimal detectable change to motor learning research - A behavioral study

Abstract

Motor learning studies face the challenge of differentiating between real changes in performance and random measurement error. While the traditional p-value-based analyses of difference (e.g., t-tests, ANOVAs) provide information on the statistical significance of a reported change in performance scores, they do not inform as to the likely cause or origin of that change, that is, the contribution of both real modifications in performance and random measurement error to the reported change. One way of differentiating between real change and random measurement error is through the utilization of the statistics of standard error of measurement (SEM) and minimal detectable change (MDC). SEM is estimated from the standard deviation of a sample of scores at baseline and a test-retest reliability index of the measurement instrument or test employed. MDC, in turn, is estimated from SEM and a degree of confidence, usually 95%. The MDC value might be regarded as the minimum amount of change that needs to be observed for it to be considered a real change, or a change to which the contribution of real modifications in performance is likely to be greater than that of random measurement error. A computer-based motor task was designed to illustrate the applicability of SEM and MDC to motor learning research. Two studies were conducted with healthy participants. Study 1 assessed the test-retest reliability of the task and Study 2 consisted in a typical motor learning study, where participants practiced the task for five consecutive days. In Study 2, the data were analyzed with a traditional p-value-based analysis of difference (ANOVA) and also with SEM and MDC. The findings showed good test-retest reliability for the task and that the p-value-based analysis alone identified statistically significant improvements in performance over time even when the observed changes could in fact have been smaller than the MDC and thereby caused mostly by random measurement error, as opposed to by learning. We suggest therefore that motor learning studies could complement their p-value-based analyses of difference with statistics such as SEM and MDC in order to inform as to the likely cause or origin of any reported changes in performance.