The CFC data were log-transformed before any of the subsequent analyses between groups due to the non-normal distribution. For comparing CFC between all three groups, we performed an ANCOVA (Gruner, 2020) with age, gender, education, and current medication as covariates. For comparing pMCI vs. sMCI based on the 2nd-year follow-up, we used a general linear model with the same covariate as mentioned above. In addition, we also compared AD and pMCI using the same technique as mentioned above. We performed correction for multiple comparisons for the electrode-to-electrode comparisons for each frequency band separately using false discovery rate. The comparisons between three groups and two groups for global CFC were computed as described above. For the correlation between gamma/theta CFC and CSF biomarkers (amyloid, total tau, and phosphorylated tau), ACE, and MMSE in patients with MCI, we used partial correlation with the same covariates as described above due to variability between subjects. The same analysis was performed to correlate ACE and total tau.
