This instrument was developed by Gross and John (2003), of self-report and fast application, consisting of 10 items [e.g. Item 03.When I want to feel less negative emotions (such as sadness or anger), I change what I am thinking about; Item 08. I control my emotions by changing the way I think about the situation I’m in.] The instrument has two modalities of emotional regulation strategies, called cognitive reassessment and emotional suppression.

The first factor, called Cognitive Reassessment, presented factorial loads varying from 0.55 (Item 5. When I’m faced with a stressful situation, I make myself think about it in a way that helps me stay calm.) to 0.83 (Item 1. When I want to feel more positive emotion (such as joy or amusement), I change what I’m thinking about.), in a total of 6 items. Next, the second factor, named Emotional Suppression, showed factorial loads ranging from 0.54 (Item 4. When I am feeling positive emotions, I am careful not to express them) to 0.83 (Item 6. I control my emotions by not expressing them.) making a total of 4 items. Together, these factors accounted for more than 50% of the total variance. Regarding the national context, the measure was linguistically adapted by Boian, Soares and Silva (2009). However, it was Bastitoni et al. (2013) who found evidence of validity of the instrument, using a national sample of 153 elderly, where it was possible to verify a factorial structure equivalent to the original study. In the national survey, the factors of Cognitive Reassessment and Emotional Suppression presented Cronbach’s alphas of 0.74 and 0.69, respectively. The factors together accounted for 50.1% of the total variance and total internal consistency (Cronbach’s alpha) of 0.73.

It is worth mentioning that this instrument has been used in different contexts internationally, showing good psychometric parameters. Eldeleklioğlu and Eroğlu (2015), seeking evidence of validity of this measure for the Turkish context, verified that it has an internal consistency index of 0.78 and 0.73 for the Cognitive Reassessment and Emotional Suppression factors, respectively. Regarding the model’s indices of adequacy to the data, the study showed CFI = 0.98, GFI = 0.99 and RMSEA = 0.06. Another example of use of the measure is in the research developed by Melka, Lancaster, Bryant and Rodriguez (2011) when studying the psychometric properties in the American context. They also observed good indicators of adequacy to the model, as can be observed by CFI = 0.96, GFI = 0.96 and RMSEA = 0.05, with internal consistency of 0.73 (Cognitive Reassessment) and 0.79 (Emotional Suppression).

In conclusion, when we take the good psychometric parameters observed in the global context into account, as well as the importance of studying strategies of emotional regulation to optimize the wellbeing, we developed this study to verify the evidences of validity of the construct and internal consistency of the Emotion Regulation Questionnaire for adolescents and young adults, since though Batistoni et al. (2011) did verify the psychometric parameters through exploratory factorial analysis, the ERQs bifactorial structure had not been confirmed.

We had 441 participants, high school students from the State of Paraíba enrolled in public (53.5%) and private (46.5%) institutions. Their ages ranged between 14 and 25 years (M= 16.1, SD = 1.14), most of them female (54.6%), single (91.6%), Catholic (46.5%), and declaring themselves of middle socioeconomic class (58.3%). This was a non-probabilistic sample (convenience), with the participation of people who, when requested, agreed to collaborate voluntarily. For the statistical analysis, expecting to have both exploratory and confirmatory analysis, we decided to divide this sample into two groups, at random, as follows: G1 (N=212), the average age was 16.1 years (SD 1.02), mostly female (51.5%), single (90.8%), catholic (45.4%) and middle- class (57.1%), and G2 (N=229), participants aged in average 16.2 years (SD=1.23), mostly female (64.2%), single (81.3%), catholic (41.8%) and middle class (59.4%).

The participants answered a booklet with sociodemographic questions [age, gender, marital status, educational institution, religion and self-perceived socioeconomic class (very low, low, medium, high and very high)] and the Emotion Regulation Questionnaire (ERQ). Elaborated by Gross and John (2003), it is composed of ten items (e.g. Item 02. I keep my emotions to myself.; Item 08. I control my emotions by changing the way I think about the situation I’m in.), answered on a seven-point Likert scale, ranging from 1 (Strongly disagree) to 7 (Strongly agree). As indicated previously, its psychometric parameters have been adequate in other cultural contexts (Cabello, Salguero, Fernández-Berrocal, & Gross, 2013; Eldeleklioğlu & Eroğlu, 2015; Yoshizu, Sekiguchi, & Amemiya, 2013), with evidences in Brazil that it works with the elderly (Batistoni et al., 2013). High scores on one of its factors indicate a greater propensity to use that strategy of emotional regulation.

Initially, this research was submitted to the Research Ethics Committee. Immediately afterwards, we contacted the heads of educational institutions to obtain consent form and to schedule the applications. With due consent, properly trained collaborators collected the data. The questionnaires were taken in a collective classroom environment, although the answers were given individually. Those who were present and who agreed to participate in the study were included in the analytical procedures. Following the ethical recommendations for research with human beings (Resolution CNS nº 510/16), the objectives, risks and benefits of the research, were clear to all, accentuating it’s voluntary and anonymous participation. We were available to clear any doubts the participants might have had. The participants of legal age ratified their collaboration by signing a Free and Informed Consent Form. In turn, the participants under the age of 18 were placed under the responsibility of the school, which sent a Statement of Assent to the guardian of each respondent. Generally, participants completed this activity in an average of 15 minutes.

We used the Student t test for independent samples to check the discriminate power of the ERQ items, seeking to evaluate the magnitude of the latent trait on the criterion groups, differentiating between participants with high and low scores (Pasquali, 2003). Then we evaluated the factorability of the data through the KMO (Kaiser-Meyer-Olkin) test, in which the recommended values are equal to or greater than 0.50 and the Bartlett’s Test which shows that the data is not an identity matrix. After adequating the data, to check the factorial structure of the construct we adopted the Principal Component Analysis (PCA) extraction method. In addition, the internal consistency was checked via Cronbach’s alpha. For these, we used the SPSS program (version 21). Subsequently, in order to proof the underlying factor structure, a confirmatory factorial analysis was performed with the AMOS statistical program (version 21). Then we used ML (Maximum Likelihood) estimator, admitting as acceptable the following adjustment indicators (Byrne, 2010; Kline, 2010; Tabachnick & Fidell, 2013): (1) χ²/df (Chi-square / Degrees of freedom) ratio, expecting values between 2 and 3, but up to 5; (2) GFI (Goodness-of-Fit Index) and AGFI (Adjusted Goodness-of-Fit Index), in which the values range from 0 to 1, however admitting those equal to or above 0.90; (3) CFI (Comparative Fit Index), also variating between 0 and 1, admitting values equal to or above 0.90 as indications of model adequacy; and, finally, (4) RMSEA (Root-Mean-Square Error of Approximation), in which the values should stay between 0.05 and 0.08, admitting those up to 0.10.Finally, the Convergent Validity was evaluated using the Average Variance Extracted (√AVE) index (≥0.50; Fornell & Larcker, 1981), as well as using Composite Reliability (CR) (≤ 0.60; Bagozzi & Yi, 1988), using the calculator made available by Gouveia and Soares (2015).

Initially, we sought to know the discriminative power of the ERQ items, considering the classical approach of the tests. In this case, internal criteria groups were established (lower and higher) from the empirical median of the total score (Md = 4.8). Then, through the t-test, we compared the means of these groups to each of the items, checking if they could differentiate scores close to the construct. The results of this analysis indicated that all items showed favorable discriminative power (5.30 < t < 9.65, p < 0.001) as described in Table 1.

Table 1

Items	Inferior Group		Superior Group		t
	M	SD	M	SD
1	4.28	1.57	5.57	1.34	6.15*
2	4.15	1.81	5.54	1.43	5.87*
3	4.27	1.61	5.71	1.36	6.73*
4	3.28	1.49	4.57	1.89	5.30*
5	4.79	1.64	6.16	1.13	6.65*
6	3.43	1.61	5.03	1.71	6.64*
7	3.99	1.43	5.60	1.20	8.47*
8	3.97	1.30	5.67	1.14	9.65*
9	4.16	1.66	5.48	1.24	5.72*
10	4.08	1.53	5.51	1.24	7.17*

* p < 0.05

Then, we verified the factorability of the matrix of correlation between the items, a condition to make a factorial analysis, proving its adequacy (KMO= 0.75 and Bartlett’s Test, χ² (45) = 457.35, p<0.001]. We then decided to perform an analysis of main components, with oblimin rotation, without fixating the number of factors to be extracted. This resulted in the identification of three components/factors with their own values equal to or greater than 1 (Kaiser Criterion): 3.04, 1.62 and 1.27, which together accounted for 59.3% of the total variance. This finding was corroborated by the parallel analysis (Horn Criterion), using the same parameters of the empirical database, performing 1,000 simulations, since only the fourth value (1.08) was higher than the observed value (0.83). The observed and simulated values are detailed in Table 2.

Table 2

Observed Values	Simulated Values	Percentiles
3.04	1.35	1.46
1.62	1.24	1.31
1.27	1.15	1.21
0.83	1.08	1.13
0.76	1.02	1.06
0.63	0.95	1.00
0.59	0.89	0.94
0.43	0.83	0.88
0.41	0.76	0.82
0.39	0.68	0.75

The factorial solution found, with the factorial loadings of the items, eigenvalues and variances explained for each factor can be verified in Table 3, which also presents the corresponding coefficients of internal consistency. We adopted as a cut-off point for the interpretation of the factor a factorial loading of 0.55, which retained most of the items; only item 9 (When I am feeling negative emotions, I make sure not to express them.) has been excluded because it has dissonant semantic content. More information about each dimension will be presented below.

Table 3

Item content	Saturation
	Factor 1	Factor 2	Factor 3
08. I control my emotions by changing the way I think about the situation I’m in.	0.74	0.44	0.18
10. When I want to feel less negative emotion, I change the way I’m thinking about the situation.	0.68	0.54	0.04
05. When I’m faced with a stressful situation, I make myself think about it in a way that helps me stay calm.	0.67	0.02	0.11
09. When I am feeling negative emotions, I make sure not to express them.	0.65	0.03	0.13
01. When I want to feel more positive emotion (such as joy or amusement), I change what I’m thinking about.	-0.03	0.78	0.18
03. When I want to feel less negative emotion (such as sadness or anger), I change what I’m thinking about.	0.23	0.76	0.01
07. When I want to feel more positive emotion, I change the way I’m thinking about the situation.	0.53	0.66	0.13
02. I keep my emotions to myself.	0.00	0.10	0.83
06. I control my emotions by not expressing them.	0.37	-0.10	0.76
04. When I am feeling positive emotions, I am careful not to express them.	0.10	0.16	0.64
Number of items	4	3	3
Own value	3.04	1.62	1.27
%Total variance explained	30.43	16.20	12.71
Cronbach’s alpha	0.67	0.63	0.64

Factor I - Cognitive Reassessment: refers to a change in the interpretation of the emotional situation. It presented an eigenvalue of 3.04, with factorial loadings between 0.74 (Item 08. I control my emotions by changing the way I think about the situation I’m in) and 0.67 (Item 05. When I’m faced with a stressful situation, I make myself think about it in a way that helps me stay calm), explaining 30.4% of the total variance; its internal consistency was 0.67.

In summary, the ERQ items were consistently discriminative, grouped into three factors / components that emerged clearly, accounting for about 3/5 of the total variance, with acceptable coefficients of internal consistency, considering the number of items (three) per factor. However, we still had to check whether this structure could be proven, confirmed, which motivated the analysis described below.

As previously suggested, although the indicators suggest evidence of construct validity and internal consistency of the ERQ, the analyzes described here were essentially exploratory. In addition, we note that the number of factors found diverged from those originally assumed. These aspects reinforce the need to evaluate the quality of the three-factor model, comparing it with the one originally observed. When comparing the model fit indices of the unifactorial model [χ² (27) = 156.24, p<0.001; χ²/df = 5.78, GFI = 0.85; AGFI = 0.76; CFI = 0.64 and RMSEA = 0.14 (IC90% = 0.123 - 0.167)], bifactorial [χ²(26) = 156.23, p<0.001; χ²/df = 6.00; GFI = 0.85; AGFI = 0.75; CFI = 0.64 and RMSEA = 0.14 (IC90% = 0.126 - 0.171)], with the trifactorial model [χ² (24) = 67.02, p<0.001; χ²/df = 2.79, GFI = 0.93; AGFI = 0.88; CFI = 0.88 and RMSEA = 0.08 (IC90% = 0.064 - 0.014)], we can see that this model is more promising. In fact, Δχ² was significant (p <0.001), corroborating that the model found, trifactorial, showed lower chi-square. All items of the instrument showed significant saturations (factorial weights, λ) statistically different from zero (λ ≠ 0; z> 1.96; p <0.05), suggesting construct validity of this measure (see Figure 1).

Figure 1

Once its factorial structure was known, its indicators of precision and convergent validity were calculated using Cronbach’s alpha, the Composite Reliability and the square root of the Average Variance Extracted (√AVE), respectively. The results found pointed to satisfactory indexes of precision and convergent validity and are presented in Table 4.

Table 4

Factor	α	CR	√AVE
Factor I. Cognitive Reassessment	0.60	0.60	0.57
Factor II. Redirecting the Focus of Attention	0.61	0.60	0.57
Factor III. Emotional Suppression	0.66	0.67	0.58

Note. α = Cronbach’s alpha CR = Composite Reliability; √AVE = Square root of the Average Variance Extracted.