2.8. Statistical Analysis

Data were analysed using Python 3.7. Data were checked for normality using the Shapiro–Wilk test and were tested for equal sphericity and variance where appropriate using Bartlett’s and Levene’s tests, respectively. The associations between continuous variables (e.g., AP-LM latency and mean neuroplastic response) were tested using linear regression models. Interaction terms were added to regression models to test for group differences in the effect of AP-LM latencies (Group*AP-LM interaction) when predicting neuroplasticity. Group means were compared using an independent samples t-test (or a Welch’s t-test where appropriate). To further examine the influence of I-wave recruitment on post-cTBS MEP amplitudes, data were analysed based on the presence or absence of late I-waves by median split on AP-LM latency, where AP-LM latency differences greater than the group median were considered to represent late I-waves, and those lower to represent early I-waves [13]. The analysis was also performed based on a split at 4 ms AP-LM difference [11]. Results were comparable, so only the median split is reported below. Data were also analysed by the direction of neuroplastic response (inhibitory, <100% baseline; or facilitatory, >100% baseline) to examine the difference in I-wave recruitment.