*2.4. Data Analysis*

SPSS 21.0 was used to perform statistical analysis on the collected data (IBM Corp., New York, NY, USA, 2012). The reliability of the instruments used was calculated using Cronbach's alpha. After checking the assumptions of normality and homoscedasticity, a multivariate analysis (MANOVA) and receiver operating characteristic curve analysis (ROC) were carried out.

## **3. Results**

Firstly, multivariate comparisons of the mean scores for the CRI-A factors were carried out by group (community/clinical) and gender, as well as for the interaction between the two variables (Table 2).

**Table 2.** Descriptive analysis for the eight CRI-A factors and univariate analysis by group (community/clinical) and gender.


LA = logical analysis; PR = positive reappraisal; SG = seeking guidance and support; PS = problem solving; CA = cognitive avoidance; AR = acceptance or resignation; SR = seeking alternative rewards; ED = emotional discharge.

> The multivariate analysis (MANOVA) gave a significant main effect by group (community/clinical), Wilks' λ = 0.901, *F* (8, 1063) = 14.594 *p* < 0.001, *η* = 0.099, and by gender, Wilks' λ = 0.982, *F* (8, 1063) = 2.372, *p* = 0.016, *η* = 0.018), although no significant main effect was found for the group–gender interaction, Wilks' λ = 0.988, *F* (8, 1063) = 1.555, *p* = 0.134, *η* = 0.012.

As for the univariate contrasts, the tests for between-individual effects (Table 2) revealed significantly higher scores in the clinical group (*p* < 0.05) than in the community group in the seeking guidance and support, problem solving, cognitive avoidance, acceptance or resignation, seeking alternative rewards and emotional discharge factors. Male participants obtained significantly higher scores (*p* = 0.014) in the logical analysis factor than female participants did.

Multivariate comparison of the average scores in the CRI-A dimensions was carried out by group (community/clinical) and gender, as well as for the interaction between the two variables (Table 3).

**Table 3.** Descriptive analysis for the four CRI-A dimensions and univariate analysis by group (community/clinical) and group-gender interaction.


The multivariate analysis (MANOVA) gave a significant main effect by group (community/clinical), Wilks' λ = 0.905, *F* (8, 1063) = 33.095 *p* < 0.001, *η* = 0.085, and by group–gender interaction, Wilks' λ = 0.990, *F*(8, 1063) = 3.430, *p* = 0.017, *η* = 0.010, although no significant main effect was found for gender, Wilks' λ = 0.998, *F* (8, 1063) = 0.593, *p* = 0.620, *η* = 0.002.

As for the univariate contrasts, the tests for between-subject effects (Table 3) show that the clinical group obtained significantly higher scores (*p* ≤ 0.001) than the community group in the cognitive, behavioural and avoidance dimensions. They also revealed a significant group–gender interaction (*p* < 0.05) in the behavioural and avoidance dimensions.

The pairwise comparisons for interaction effects showed, on the one hand, that the difference between the groups (clinical/community) in the behavioural dimension was only significant (*p* = 0.002) between men, and, on the other hand, that men obtained significantly higher scores in the behavioural (*p* = 0.013) and avoidance (*p* ≤ 0.001) dimensions in the community sample.

To analyse differences in the most commonly adopted strategies between the community and clinical groups, a repeated-measures ANOVA was performed between the scores for the cognitive, behavioural, approach and avoidance dimensions for the community and clinical groups (Table 4).



The repeated-measures analysis showed significant differences (*p* < 0.05) between the scores obtained in the four dimensions of the CRI-A in both groups (Table 4). The pairwise comparisons indicated the following: (1) In the community group, significantly higher scores (*p* < 0.001) were obtained in the cognitive and approach dimensions than in the behavioural and avoidance dimensions, with the score (*p* < 0.001) for the approach dimension being significantly higher than for the cognitive dimension. (2) In the clinical group, significantly higher scores (*p* < 0.05) were obtained in the cognitive and avoidance dimensions than in the behavioural and approach dimensions.

In order to evaluate the discriminative accuracy of the coping responses, a receiver operating characteristic curve analysis (ROC) was carried out to identify the cut-off points for the scores in the dimensions of the CRI-A (cognitive, behavioural, approach, avoidance), after which behavioural disorders become more likely.

In the nonparametric ROC analysis (Figure 1), the area beneath the curve for the cognitive dimension is 0.609 (*p* < 0.001; confidence interval 95%; min. = 0.555; max. = 0.622), the area beneath the curve for the behavioural dimension is 0.582 (*p* = 0.002; confidence interval 95%; min. = 0.529; max. = 0.635), the area beneath the curve for the approach dimension is 0.496 (*p* = 0.875; confidence interval 95%; min. = 0.444; max. = 0.547) and provides no significant information, and finally, the area beneath the curve for the avoidance dimension is 0.724 (*p* < 0.001; confidence interval 95%; min. = 0.683; max. = 0.765).

**Figure 1.** ROC curve for the CRI-A dimensions predicting the presence of psychopathological disorders.

Table 5 shows the cut-off points maximising both sensitivity and specificity and the cutoff points maximising sensitivity and specificity for the dimensions providing significant information (*p* ≤ 0.05).

With regard to the presence of psychopathological disorders, a score of ≥42.5 in the cognitive dimension maximised both sensitivity (58%) and specificity (58%) (Youden's index = 0.159), a score of 41.5 maximised sensitivity (60%) while maintaining specificity higher than random, and a score of 43 maximised specificity (60%) while maintaining sensitivity higher than random. In the behavioural dimension, a score of ≥39.5 maximised both sensitivity (53%) and specificity (59%) (Youden's index = 0.124), a score of 37.5 maximised sensitivity (56%) while maintaining specificity higher than random, and a score of 40.0 maximised specificity (61%) while maintaining sensitivity higher than random. Finally, in the avoidance dimension, a score of ≥39.5 maximised both sensitivity (64%) and specificity (65%) (Youden's index = 0.293), a score of 36.0 maximised sensitivity (79%) while maintaining specificity higher than random, and a score of 42.0 maximised specificity (73%) while maintaining sensitivity higher than random.


**Table 5.** Sensitivity, specificity and Youden's index for scores in the cognitive, behavioural and avoidance dimensions of the CRI-A to identify the presence of psychopathological disorders.

\* Score maximising sensitivity. \*\* Score maximising specificity. \*\*\* Score maximising sensitivity and specificity.
