*2.5. Statistical Analysis*

Statistical analysis was performed with Stata17 for Windows [63]. Comparison between the groups was performed with analysis of covariance (ANCOVA, adjusted for sex, age, educational level, and BMI) for quantitative measures and with logistic regression (also adjusted for the same covariates) for binary measures. Since the groups were ordered according to their position within the compulsivity–impulsivity continuous dimensional spectrum, these models included polynomial contrasts to assess the presence of patterns in data adjusted by linear, quadratic, cubic, and quartic equation/functions (four polynomial trends were assessed, the maximum allowed for variables categorized in five group levels). Pairwise post hoc comparisons also explored differences between group means and proportions.

The Finner method was employed to control type I error due to multiple null hypothesis tests. This is a correction procedure based on a stepwise multiple-test method aimed to adjust *p*-values whilst controlling the familywise error rate (FWER, defined as the likelihood of achieving at least k false rejections) [64]. Controlling the *k-*FWER implies fixing a number of *k*−1 of tolerated erroneous rejections, and then combining the unadjusted *p*-values to obtain a single testing for the group of null hypothesis tests at α-level.
