Data Analysis

The first ten taps were excluded to allow performance during synchronization to stabilize. The raw data were examined to screen for places in which a tap was missing or an extra response was recorded, and only intact segments of each trial were used for parameter estimation. Trials in which the asynchronies showed sudden drifts of the mean (for example, to an antiphase pattern) were excluded. These exclusions were far more frequent for the two patient groups (16%) than for the controls (3%). Parameter estimates from trials in which the autocorrelation function was degenerative, yielding indeterminacy of the model, were also excluded (9%). These trials were frequent for both the control and patient groups. Another effect to be considered is inconsistency between runs. Given the likelihood of stochastic variations in process control parameters, moderate effects of this kind are normative and can be addressed by averaging across runs. However, it is possible that performance changes over time due to learning. For example, with practice, Pressing (1999) observed an increase in the utilization of error correction and a decrease in the estimate of central variability. However, these effects occurred over a period of years. We assume that learning-related changes are minimal with the current design.

0 0

Post a comment