A Comparison of Statistical Models for Calculating Reliability of the Hoffmann Reflex

The Hoffmann reflex is obtained through surface electromyographic recordings, and it is one of the most common neurophysiological techniques in exercise science. Measurement and evaluation of the peak-to-peak amplitude of the Hoffmann reflex has been guided by the observation that it is a variable response that requires multiple trials to obtain a stable mean. As a result, reliability assessment of the Hoffmann reflex has thus far been limited to trial-to-trial variability or the mean of multiple trials across test days to evaluate day-to-day variability, without considering how the two dimensions of measurement interact to affect test–retest reliability. This article compares two analysis of variance models for reliability assessment of the peak-to-peak amplitude of the Hoffmann reflex: one model incorporated both days and trials, and the other used the mean of multiple trials on each test day. The intraclass correlation coefficient calculated from an analysis of variance model that incorporated both trial-to-trial and day-to-day error variances was R = .93. Using the mean of the trials for each test day as a single score resulted in a decrease in the intraclass correlation coefficient (R _2,1 = .76). This study demonstrated that an analysis of variance model that accounts for both trial-to-trial and day-to-day error variances provided a higher estimate of reliability. Further, reliability estimation based on the interaction between both trials and days revealed that a schedule with additional test days had a disproportionate impact upon securing a reliable measure of Hoffmann reflex amplitude.