*2.8. Statistical Analysis*

Descriptive statistics were computed for all variables in the analysis. Independent sample *t*-tests assessed for gender differences. The prevalence of alexithymia was determined as a percentage. The internal consistency reliability of the BAQ, SOC-13 and TAS-20 in our sample was evaluated using Cronbach's alpha coefficient (≥0.70). The Shapiro–Wilk test was used to assess the normality of the data. Pearson correlation was performed to determine the strength and direction of the relationship between variables. Hierarchical linear regression analyses were built to investigate whether related variables were significant predictors of aggression while controlling for other covariables. The assumption testing (linear relationship, independence, homoscedasticity and normality) was carried out by visual inspection of the variables, residuals and collinearity statistics and quantile–quantile (QQ) plots, probability–probability (PP) plots and scatterplots. A bootstrap approach was used to test the significance of the indirect effect of alexithymia on aggression through the mediating role of sense of coherence. The SPSS PROCESS Macro (Hayes, 2013) was used to conduct simple mediation analyses, computing 5000 bootstrap resampling with replacement from the original dataset to estimate 95% confidence intervals (CIs) for the indirect effects (CIs that do not include zero indicate a significant indirect effect). For the sake of parsimony, mediation models were run, including TAS-20 and SOC subscales, predicting each of the BAQ subscales. Path analysis was performed in order to concurrently examine the impact of a set of predictor variables (certain TAS-20 and SOC subscales derived from regression and mediation model results) on the BAQ subscales, which were handled as dependent variables and thus identify which are the most important (and significant) paths. This may have implications for the plausibility of our prespecified hypotheses. A structural model with observed variables was tested using a covariance matrix as input and maximum likelihood estimation. Assumptions (linearity, causal closure, unitary variables) were respected. Maximum likelihood estimation (MLE) indices were calculated in order to assess the correspondence of the model with the data: chi-square statistics, root mean square error of approximation (RMSEA) and comparative fit index (CFI). All *p* values were two-tailed, and the statistical significance level was set at *p* < 0.05. SPSS software, version

23, was used for the statistical analysis. SPSS AMOS 23 Graphics enabled the presentation of Figures 1–5.

**Figure 1.** Simple mediation analysis of sense of coherence (SOC) on Toronto alexithymia scale (TAS)–brief aggression questionnaire (BAQ) relationship.

**Figure 2.** Simple mediation analysis of difficulty identifying feelings (DIF) on hostility (H)–anger (A) relationship.

**Figure 3.** Simple mediation analysis of comprehensibility (SOC B) on hostility (H)–physical aggression (PA) relationship.

**Figure 4.** Simple mediation analysis of manageability (SOC C) on anger (A)–physical aggression (PA) relationship.

**Figure 5.** Path model illustrating patterns of effect within a system of research variables. Note: Standardized coefficients are presented.

#### **3. Results**

*3.1. General Characteristics of Participants and Scores on Outcome Variables*

The study included 100 participants (45 men and 55 women). Means and standard deviations for general characteristics of participants and all key variables are presented in Table 1. The mean BAQ score was statistically higher compared to the corresponding score in the Greek general population [107], (30.93 vs. 23.22, sample *t*-test *p* < 0.01). A total of 22% of schizophrenic participants scored above the cutoff on the TAS scale. The mean TAS score was statistically higher compared to the corresponding score in the general population [61], (49.10 vs. 45.8, sample *t*-test *p* < 0.05), but comparable to the average score observed among patients with chronic somatic diseases (49.10 vs. 48.2 sample *t*-test *p* > 0.05) [108]. The average SOC was statistically lower compared to the mean score from the standardization studies [104] in the Greek general population (56.57 vs. 59.85 sample *t*-test *p* < 0.05). No statistically significant differences were observed between men and women as to the BAQ, TAS, or SOC scores or any of their subscales. Significantly higher BAQ and lower SOC scores were evidenced among alexithymic participants (36.59 ± 8.7 vs. 29.33 ± 7.42, t = 3.564, *p* = 0.001 and 45.95 ± 16.33 vs. 59.42 ± 1433, t = −3.44, *p* = 0.002, correspondingly).


**Table 1.** General characteristics of participants and scores on BAQ, TAS, SOC and subscales.

Abbreviations: P, participants; D.S., descriptive statistics; PA, physical aggression; VA, verbal aggression; H, hostility; A, anger; DIF, difficulty identifying feelings; DDF, difficulty describing feelings; EOT, externally oriented thinking; SOC A; meaningfulness; SOC B, comprehensibility; SOC C, manageability.

#### *3.2. Correlations among Continuous Variables*

Significant negative correlations were evidenced among scores on the SOC scale (*p* < 0.001) with both the TAS as well as with BAQ scales. However, a positive correlation (*p* < 0.001) was indicated between TAS and BAQ scales (Table 2).

Results from correlations among subscales of the TAS, BAQ and SOC are presented in Table 3.


**Table 2.** Correlations among age, education (in years), TAS, BAQ and SOC.

\*\* *p* < 0.01.

**Table 3.** Correlations among subscales of TAS, BAQ and SOC.


Abbreviations: PA, physical aggression; VA, verbal aggression; H, hostility; A, anger; DIF, difficulty identifying feelings; DDF, difficulty describing feelings; EOT, externally oriented thinking; SOC A; meaningfulness; SOC B, comprehensibility; SOC C, manageability. \* *p* < 0.05 or \*\* *p* < 0.01.
