*2.4. Statistical Analysis*

Statistical analysis was carried out with Stata17 for Windows (Stata-Corp, College Station, TX, USA) [29]. Firstly, the empirical clusters compared in this study were compared for the measures assessed at baseline, and for the ED diagnostic subtype, to provide information about the specific profile associated to each cluster. Next, the discriminative capacity of the empirical clusters on the treatment response was assessed.

Comparison between the clusters was based on chi-square tests (χ2) for categorical variables and with analysis of variance (ANOVA) for quantitative variables. Effect size was calculated with the standardized Cohen-*h* for proportion differences and Cohen-*d* for mean differences (poor effect size was considered for absolute estimates lower than 0.50, mild-moderate effect size for absolute estimates higher than 0.50 and high-large for absolute estimates higher than 0.80) [30]. Control in Type-I error due to the multiple null-hypothesis statistical tests was done with Finner's method [31].
