«Development and structural validation of a scale to assess regulation of anger and sadness in interpersonal situations* M. Florencia Giuliani1 ...»
Eight strategies were taken into account: cognitive reappraisal (Cabello, Salguero, Fernández Berrocal, & Gross, 2013), expressive suppression (Cabello et al., 2013), emotional repair (Fernández-Berrocal, Extremera, & Ramos, 2004), seeking emotional support (Perczek, Carver, Price, & Pozo-Kaderman, 2000), situation modification (Perczek et al., 2000), selection of situations (Carstensen, Fung, & Charles, 2003), attentional deployment (Urry & Gross, 2010), and acceptance (Cebolla, García, Soler, Guillen, Baños, & Botella, 2012).
A set of sentences based on the definition of each strategy was created, including some items extracted from existing instruments. A panel of three independent experts in the field judged the adequacy and clarity of the items. After reviewing the panel’s suggestions 62 items were conserved and randomly ordered in the scale. Six items belonged to the strategy “selection of situations” and eight items to each of the other strategies considered.
Two versions of the instrument were developed, one for the anger situation and the other for the sadness situation. Items were identical in both scales, and changes were only introduced when the emotion was explicitly mentioned in the sentence (e.g., “I think feeling sad is understandable, I wouldn’t do anything to change it” and “I think feeling angry is understandable, I wouldn’t do anything to change it”).
Tutors on the undergraduate study programs previously mentioned were approached by the first author of this study. After explaining the objectives of the study and the procedure for data gathering, she sought their agreement to participate.
Data were gathered by means of a self-administered questionnaire. The first page consisted of a statement regarding voluntary participation in the study and the confidential and anonymous use of data. The questionnaire also included brief information on the goals of the research and an informed consent document which participants were asked to sign. Only questionnaires that were returned with the consent form signed were included in the final sample.
The questionnaire was applied collectively in university lecture rooms but was answered individually. Of the total number of participants, 234 answered the anger version of the scale and 166 the sadness version. Doubts arose only exceptionally and were clarified orally by the researcher. The administration took a maximum of 20 minutes.
An analysis of missing data showed that in no case did this represent more than 5% of the data. Missing data were then estimated using the multiple imputations method (Kline, 2011). An exploratory factor analysis (EFA) was performed with the results of the scale for 1) the anger situation (Bartlett’s test of sphericity χ2(1891) = 8984, p.001; KMO = 0.84), 2) the sadness situation (Bartlett’s test of sphericity χ2(1891) = 6549, p.001; KMO = 0.77), and 3) both situations considered altogether (Bartlett’s test of sphericity χ2(1891) = 13960, p.001; KMO = 0.87). In light of these results, subsequent analyses were performed with data derived from the EFA for both situations considered altogether. The EFA´s results were similar, with the exception of three items whose factor loadings differed between the anger and sadness situations and which were therefore excluded from the final item pool.
The EFA used the maximum likelihood method and Promax rotation. Given the potentially high correlations among the factors an oblique rotation was applied (Ferrando & Anguiano-Carrasco, 2010). In order to select the optimal number of factors a parallel analysis was carried out (Pérez & Medrano, 2010). As a result, seven of the eight factors considered were retained, all with eigenvalues higher than 1.
In order to create a shorter version of the scale with adequate psychometric properties, four items were selected from each dimension. This selection was made on the basis of three criteria: 1) The EFA results, retaining items with higher loadings on their corresponding factor and lower loadings on other factors; 2) the Cronbach’s alpha index; and 3) the item content, prioritizing items that were not repetitive. These criteria were applied in all but two cases. First, in the case of Expressive Suppression and Seeking Emotional Support, items loading positively on the first strategy presented a negative loading on the second, and vice versa.
Second, some items corresponding to Cognitive Reappraisal presented moderate loadings on Situation Modification. In order to maximize the independence of factors, items meeting the abovementioned criteria and presenting lower loadings on other factors were retained. Table 1 shows the items selected for the final version of the scale, along with their factor loadings and the percentage of variance accounted by each factor. The table also includes Cronbach’s alpha values, which were adequate and ranged from.75 to.87.
Anuario de Psicología/The UB Journal of Psychology, vol. 45, nº 1, abril 2015, pp. 115-130 © 2015, Universitat de Barcelona, Facultat de Psicologia
TABLE 1. EXPLORATORY FACTOR ANALYSIS WITH THE MAXIMUM LIKELIHOOD METHOD FOR THE 28-ITEM VERSION OF THE EMOTION REGULATION SCALE,
CONSIDERING BOTH SITUATIONS (ANGER AND SADNESS) TOGETHER. THE FIRST COLUMN SHOWS THE ORDER OF ITEMS IN THE SCALE
(ITEM TRANSLATION INTO ENGLISH IS PROVIDED FOR HELPING NON-SPANISH SPEAKERS TO UNDERSTAND WHAT THE MEANING OF EACH ITEM IS).
Note. The table only includes items with loadings of at least.30. Values in italics indicate item loads on that factor. SE: Seeking Emotional Support, ES: Expressive Suppression, AD:
Attentional Deployment, SM: Situation Modification, AC: Acceptance, CR: Cognitive Reappraisal, SS: Selection of Situations.
124 Emotional regulation in interpersonal situations Regarding the correlation between dimensions, positive and moderate correlations were found between Situation Modification and Cognitive Reappraisal (r =.32, p.01), Situation Modification and Selection of Situations (r =.32, p.01), and Selection of Situations and Attentional Deployment (r =.30, p.01). In addition, a negative and moderate correlation was found between Seeking Emotional Support and Expressive Suppression (r = -.48, p.01).
The scale developed was called as Scale of Emotion Regulation in Interpersonal Situations (SERIS). In Study 2 we tested the scale resulting from Study 1 using confirmatory factor analysis (CFA) with a sample of similar characteristics.
Study 2 Method Participants This study used a convenience sample of 259 first-year, second-year, and third-year psychology undergraduates from the National University of Mar del Plata (Argentina). The mean age of the sample was 22.2 years (SD = 5.3), and there were 197 females and 60 males (2 students did not indicate their gender).
Instruments and Procedure
A questionnaire with two sections was administered to participants. Section 1 covered sociodemographic information and Section 2 included the scale for assessing the ER of anger and sadness in the context of family ties (SERIS), developed in Study 1.
The procedure used in this study was identical to that used in Study 1. Of the total number of participants, 130 answered the anger version of the scale and 129 the sadness version. The administration took about 10 minutes.
The missing data analysis showed that in no case did this represent more than 5% of the data. Thus, as in the previous study, missing data were estimated using the multiple imputations method (Kline, 2011). We then calculated values of Cronbach’s alpha and correlations between dimensions. Internal consistency coefficients ranged from.73 (Situation Modification) to.87 (Seeking Emotional Support). Correlations between dimensions were lower than.30, with the exception of
those between Cognitive Reappraisal and Attentional Deployment (r =.31, p.01), Expressive Suppression and Seeking Emotional Support (r = -.43, p.01), Selection of Situations and Situation Modification (r =.45, p.01), and Selection of Situations and Attentional Deployment (r =.38, p.01).
*p,05 **p,01 Finally, a confirmatory factor analysis (CFA) was conducted using SPSS AMOS 21. As a previous step, the data distribution was analyzed. Results showed
that skewness and kurtosis did not deviate from a normal distribution. Three factor analytic models were then compared:
1. A seven independent factors structure, as suggested by the results of the EFA, in which the factors were permitted to correlate between them.
2. A seven-factor structure with a global second-order latent factor, called the ER factor.
3. A seven-factor structure with two second-order latent factors, one encompassing the antecedent-focused strategies (Cognitive Reappraisal, Situation Modification, Attention Deployment, and Selection of Situations) and the other the response-focused strategies (Seeking Emotional Support, ExpresAnuario de Psicología/The UB Journal of Psychology, vol. 45, nº 1, abril 2015, pp. 115-130 © 2015, Universitat de Barcelona, Facultat de Psicologia 126 Emotional regulation in interpersonal situations
sive Suppression, and Acceptance), as in the model proposed by Gross (2014).
In order to explore the factor structure of the scale the fit of the different models was assessed using the maximum likelihood estimation technique. Chisquare values and goodness-of-fit indexes were evaluated following the criteria established by Hu and Bentler (1998).
The results showed that none of the models fitted the values indicated by the literature (Hu & Bentler, 1998). However, of the three models the seven independent factors model yielded better goodness-of-fit indexes. Consequently, we tested a new model generated from a revision of the modification indexes (MI;
Kline, 2011) of this first model. Modification indexes estimate the amount by which the fit between the data and the model would improve if the analysis were repeated, eliminating the restrictions applied to items. High MI values suggest that items may have a common source of error that is not represented in the model. The standard procedure to improve the goodness-of-fit of the model was followed, consisting in including a correlation between the errors of the items, provided that items belonged to the same factor. Items with MI values higher than 13 (the highest found) and belonging to the same dimension were modified. Thus, the errors of items 7 and 8 (Acceptance), 11 and 12 (Situation Modification), and 25 and 26 (Cognitive Reappraisal) were correlated. This procedure means that CFA becomes a multidimensional measurement model, in other words, a nonstandard CFA (Kline, 2011). The new model showed more appropriate GFI, CFI, and RMSEA values than did the previous ones (see table 3).
Invariance of configuration was assessed using R Software. Model 1 (modified) was used for both subsamples (anger and sadness) to validate the structural equivalence of the instrument in both scenarios. The tests examined if the scale measure properties are similar for both groups (Brown, 2012). A specific proce
dure for small samples (ratio equal or less than one participant for each two estimated parameters) was applied, consisting of a correction of the maximum likelihood chi-square statistic for the estimation of noncentrality-based fit measure, as proposed by Swain (Herzog & Boosma, 2008). Results suggested an adequate
goodness of fit, both for the sadness situation (CFI:.87, TLI:.88 and RMSEA:
.06) and for the anger situation (CFI:.89, TLI:.88 and RMSEA:.05).
Finally, we compared the frequency of use of ER strategies in both situations. Results showed that seeking emotional support and attentional deployment were more frequent in the sadness situation, whereas situation modification and selection of situations were more frequent in the anger situation. Acceptance, cognitive reappraisal, and expressive suppression showed no significant differences (see table 4).
TABLE 4. MEANS, STANDARD DEVIATIONS, AND ‘T’ VALUES OF ER STRATEGIES
WITH RESPECT OF THE ANGER SITUATION AND THE SADNESS SITUATION.
Discussion This paper describes the development and validation of a new instrument to assess ER, the SERIS. This instrument considers a full range of strategies, takes into account the relational context of emotions, and explicitly mentions the emotion in question. This instrument is aimed at improving the ecological validity of ER assessment (Aldao, 2013) while keeping a degree of standardization that allows for interindividual and intergroup comparisons.
The results of Study 1 suggest that items created from previous literature on ER were empirically consistent in both the anger and sadness situations. In fact, the dimensions identified by the EFA were the same in both situations. The dimensions of Situation Modification, Cognitive Reappraisal, Selection of SituaAnuario de Psicología/The UB Journal of Psychology, vol. 45, nº 1, abril 2015, pp. 115-130 © 2015, Universitat de Barcelona, Facultat de Psicologia 128 Emotional regulation in interpersonal situations tions, and Attentional Deployment were moderately and positively correlated.
This is consistent with the modal model of emotions (Gross, 2014), which asserts that all these strategies can be considered as antecedent-focused strategies.
Study 2 was aimed at validating the scale’s factor structure identified in the EFA. The final version of the instrument has 28-item, organized in 7 independent factors, which is a common structure in instruments measuring ER from a multidimensional perspective (Gross & John, 2003). Internal consistency coefficients were adequate in both studies, and correlations between sub-scales were also similar.