*2.4. Data Analyses*

Descriptive statistics were used to determine the frequency rates of online hate. Pearson's r correlations were used to investigate the bivariate associations among the main study's variables. The *t*-test was used to investigate sex differences among the online hate variables, Cohen's *d* used to calculate the effect size. Confirmatory Factor Analysis was completed with Mplus 8.1 software (Muthén & Muthén, Los Angeles, CA, USA) [26]. The proposed regression-based moderated model was examined using the Process Macro for SPSS (SPSS Inc., Chicago, IL, USA) [27], applying Model 1 with 5000 bias-corrected bootstrap samples. Being bystanders of online hate was the independent variable, toxic online disinhibition was the moderator, and online hate perpetration was the dependent variable, while controlling for participants' age, sex, migration background, socioeconomic background, and online hate victimization. Cohen's *f* 2 was used as an effect size. According to Cohen [28] *f* 2 ≥ 0.10, *f* 2 ≥ 0.25, and *f* 2 ≥ 0.40 represent small, medium, and large effect sizes, respectively. Multicollinearity diagnostics were assessed and were within an acceptable range (see Table 1).


**Table 1.** Means, standard deviations, and correlations between online hate bystanders, online hate perpetrators, online hate victims, and toxic online disinhibition.

> \**p* < 0.05; \*\* *p* < 0.01.
