1. Introduction
In recent times, the study of emotions in the university context versus academic achievement has been a subject of great interest (Lasarte et al., 2019; Pekrun et al., 2002; Supervía et al., 2020; Vera & Morales, 2019). Even though there is an extensive literature that involves the study of various psychological factors related to university performance (Martínez-Monteagudo, 2019; Parasi, 2019; Toro, 2019), these have denoted the undeniable role of emotions in the academic particularities that the student is studying (González et al., 2009; Pekrun et al., 2000; Turner & Schallert, 2001).
Regarding the construct of emotional regulation, it is considered that students experience a wide repertoire of emotions during their learning experiences at university (Medrano & Trógolo, 2014), such as manifested emotions of rejoicing, illusion, conceit, consolation, anger, distress, discouragement, and shyness (Pekrun et al., 2002). Therefore, the relationship between emotional regulation and university development is complex, managing to determine what students do academically (Diego et al., 2018; Medrano & Trógolo, 2014).
Regarding the theoretical description of the construct, Thompson defines emotional regulation as “the set of extrinsic and intrinsic processes responsible for monitoring, evaluating and modifying emotional reactions, especially their intensity and characteristics over time, with the aim of achieving the own goals” (1994, pp. 27-28). However, Gross (1998) exposes emotional regulation as “the process by which individuals order their emotions, how they experience and express them” (p. 275). These arguments establish that emotional regulation is a mechanism by which emotional expression is harmonized in or-der to achieve an explicit objective. On the other hand, regarding the psychopathological representation of the construct, various investigations mention that difficulties in emotional regulation are associated with certain characteristic mechanisms in mental disorders (C. Silva, 2005), such as anxiety and depression (Mennin et al., 2007), additions (McNally et al., 2003), eating behavior problems (Pascual et al., 2011; Serppe et al., 2018; J. R. Silva, 2008), schizophrenic outbreaks (Kring & Neale, 1996), personality disorders (Glenn & Klonsky, 2009; Gratz & Roemer, 2008), among others. However, considering the measurement of the construct, significant relevance has been denoted on measurement questionnaires for emotional regulation (Kinkead et al., 2011), ranging from a structured interview to self-applied scalar tests (Caprara & Pastorelli, 1993; Gálvez-Hernández et al., 2019; Garnefski et al., 2001; Gratz & Roemer, 2004; Greenwald et al., 2003; Gross & John, 2003; Hoshmand & Austin, 1987; Kamholz et al., 2006; Katz et al., 1999; Lane et al., 1990; Larsen & Prizmic, 2004; Rovira et al., 2012; Schutz et al., 2004; Stanton et al., 2000).
According to this wide variety of instruments, it is considered that the Scale of difficulties in emotional regulation (DERS; Gratz & Roemer, 2004) manages to denote an influx in both its composition and its use in research and clinical practice. It has recently adapted to the Latin American context in countries such as Chile (Guzmán-González et al., 2014), Colombia (Muñoz-Martínez et al., 2016), and Argentina (Medrano & Trógolo, 2014), showing good properties on Psychometric validity and reliability, ranging from .72 to .85 of Cronbach’s alpha coefficient, and denoting it as “good” in terms of fit and reliability.
DERS was initially developed to be used in the field of clinical psychology; however, subsequent studies high-light its usefulness in non-clinical populations, as is the case of the university student population (Hervás & Jódar, 2008; Tejeda et al., 2012). According to Gratz and Roemer (2004), it is not possible to evaluate and understand emotional regulation without considering the specific context in which the situational demands and individual goals are developed.
Despite the wide variety of instruments to assess emotional regulation in various contexts, a standardized measuring instrument to assess this psychological notion is needed in Ecuador; therefore, this research corresponds to an instrumental study, where the psychometric proper-ties are examined by DERS in the Ecuadorian university population, concerning the purpose of study, to develop the translation of DERS into the Ecuadorian cultural dialect, as well as to study its structure and internal consistency of the scale in the Ecuadorian population.
2. Method
2.1 Design
The present investigation of instrumental design considered three studies: the first was to develop the translation of the original version of the scale into the Ecuadorian dialect with the translation-back-translation process; the second was to explore the underlying factors of the scale using Exploratory Factor Analysis; and the third study was to confirm the factorial structure found and submit it to a Confirmatory Factor Analysis, as well as to analyze the reliability of the instrument.
2.2 Participants
The first study had a homogeneous target sample of n= 29 university students. However, the following two studies consisted of n= 586 participants each, corresponding to an accidental non-probabilistic sample (Hernández Sampieri et al., 2015), with a total universe of n=1172 participants from 17 to 32 years of age (M = 21.99; DT = 2.49), which 58.6% were women and 41.4% were men. All of them were undergraduate students of psychology at the Catholic University of Cuenca in Ecuador. Regarding the selection criteria, for the inclusion criterion, university psychology students who signed the informed consent were considered, and participants who were under the influence of narcotic drugs, except tobacco, were excluded as part of the criteria.
2.3 Instruments
Difficulties in Emotion Regulation Scale -DERS- (Gratz & Roemer, 2004) is a self-applied questionnaire, which allows to measure the difficulties in emotional regulation and different aspects of it. Composed of 36 items measured on a five-point Likert scale (from “Almost never/0-10% of the time” to “Almost always/90-100% of the time”). Regarding its reliability and internal composition, the scale is made up of the dimensions: Limited access to regulation strategies Strategies α=.88 (8 items), Impulse control difficulties Impulse α=.86 (6 items), Interference in Goals-directed behaviors α=.89 (5 items), Lack of emotional acceptance Nonacceptance α=.85 (6 items), Lack of emotional awareness Awareness α=.80 (6 items), and Lack of emotional clarity Clarity α=.84 (5 items). In addition, the instrument has predictive validity and a high total internal consistency (α=.93). The interpretation is made with the sum of all of its items or with the specific sum of each subscale.
Sociodemographic questionnaire. A short survey that collects personal data such as age, gender, and level of study was prepared.
2.4 Process
As for the first study, the scale translation was developed in order to linguistically adapt the instrument, following the required standards (Beaton et al., 2000; Bullinger et al., 1998; Muñiz et al., 2013; Streiner & Norman, 2008) of the International Commission for Evaluation Tests (Hambleton & Kanjee, 1995; Hambleton & Bollwark, 1991). Therefore, this required four experts in bilingual translation (Spanish-English). The first couple translated autonomously the original questionnaire into Spanish, originating two translated documents. Subsequently, the second pair of translators carried out the back translation of both documents into the original language (English) of the scale (Harkness & Schoua-Glusberg, 1998; Streiner & Norman, 2008). The successive translated documents were studied and com-pared by two experts in psychometrics, denoting their semantic equality. Due to this, a version adapted to the Ecuadorian dialect was obtained, with which, from a sample (n=29) made up of Ecuadorian university students, a pilot test was carried out in order to know its feasibility (Argimon & Jiménez, 2013; García et al., 2009). After that, its semantic understanding was ensured, considering the final version of the scale.
Subsequently, the Department of Ethics of the Catholic University of Cuenca certified both the research and the informed consent, which under the APA code of ethics for research and data confidentiality (APA,1994), students who wanted to join the second and third research study proceeded voluntarily to sign the document in writing.
Both for the analysis of the second and third studies, the data collection process was carried out during teaching hours, highlighting the anonymous nature of the information collected (Behnke, 2006), for which the Ecuadorian linguistic version of DERS determined in the first study, combined with sociodemographic information. After that, the analysis of the psychometric properties of the scale was carried out transversally (Ramada-Rodilla et al., 2013; Streiner & Norman, 2008), in order to have a reliable measurement instrument (Carvajal et al., 2011).
2.5 Analysis of data
In the second and third study, the description of the sociodemographic variables was performed, as well as the univariate and multivariate normality analysis (Mardia Coefficient) on the matrix of the observed variables data.
Subsequently, a cross-validation procedure was per-formed in order to know the internal composition of fac-tors determined in DERS.
In the second study, corresponding to the exploratory factor analysis, the convenience of this analysis was assessed in advance using the KMO test (Kaiser-Meyer-Olkin) and the Bartlett sphericity test, and the same method by Gratz and Roemer (2004), of factor extraction by factorization of main axes, was used. However, given the relevance of the data, a type of oblique rotation (Promax) was used.
Concerning the third study, a confirmatory factor analysis was developed in order to demonstrate whether the structure achieved by main axes was replicated. Ac-cording to the breach of the assumption of multivariate normality in the data matrix, the statistical estimation procedure was Unweighted least squares (ULS), an estimation procedure that provides appropriate evaluations, without the requirement of the assumption of normality in the data (Ruiz, 2000). The fit of the model was evaluated by managing several values, some of which are the Goodness of Fit Index (GFI), Normed Fit Index (NFI), and the Adjusted Goodness of Fit Index (AGFI), whose values, equal or greater than .90, are deciphered as values corresponding to an admissible fit of the model (Hu & Bentler, 1999). In addition to this, the Residual Mean Square Root (RMR) was also obtained, of which a small value is interpreted as a good model (Tabachnick & Fi-dell, 2007).
Once the factorial structure was established, the reliability of the scale was known, using the Cronbach’s al-pha (α) coefficient (Nunnally, 1975) and the McDonald (ω) Omega coefficient 1999. Despite the classic use of Cronbach’s alpha coefficient, the analysis of the Omega coefficient was incorporated into the study, a relatively new reliability estimator used in factor models (Ventura- León, 2018) that manages to present a high sensitivity compared to other estimators (Zinbarg et al., 2005), as well as its robustness when sampling heterogeneous populations, the reduced risk of overestimating reliability (Waller, 2008), and its ability for not having correlated errors, which are often limitations of Cronbach’s alpha (Dunn et al., 2014; Ventura-León, 2018). Acceptable reliabity values above .70 were considered in the present study (Nunnally & Bernstein, 1994).
Second study | Third Study | χ2 | df | p | |||||
n | % | n | % | ||||||
Gender | .088 | 1 | .767 | ||||||
Female | 341 | 58.20 | 346 | 59 | |||||
Male | 245 | 41.80 | 240 | 41 | |||||
x | SD | x | SD | t | df | p | |||
Age | 21.07 | 2.49 | 21.52 | 2.47 | -3.073 | 1170 | .002 | ||
Mardia’s Coefficient | χ2 | df | p | ||||||
Am | 177.541 | - | 17433.454 | 8436 | .000 | ||||
Km | 1701.768 | 5965.014 | 1 | .000 | |||||
Am | - | 274.034 | 26908.420 | 8436 | .000 | ||||
Km | 1626.561 | 3579.713 | 1 | .000 |
Note. The data in bold correspond to the level of significance (level .05).
Statistical analyzes were performed with computer programs: SPSS software vers. 24.0.0 combined with the AMOS 24 statistical package (IBM Corp., 2016) and the JASP 0.9.2 software (Love et al., 2015).
3. Results
3.1 First study
Corresponding to the first study, as for the translation, it can be mentioned that of the 36 original items on the scale, 33 showed no problems and were translated liter-ally. In the three remaining items, minimal changes had to be made, without modifying the meaning or structure of the item. Thus, the phrase “My emotions feel overwhelming”, which describes a specific environment, was replaced by “I feel my emotions as stunners”, which hints at a more neutral state. As for the phrase “I experience my emotions as overwhelming and out of control”, it was rephrased as “I experience my emotions as stunning and out of control”. As for the statement “When I’m upset, I believe that wallowing in it is all I can do”, it was modified to the phrase “When I’m up-set, I think that sinking into it is all I can do”. The pilot test (n=35) confirmed the congruence of the scale. Finally, as regards the understanding of the items, all the university students confirmed the feasibility of the instrument with clear and simple language.
In both the second and third studies, in the cross-validation process, the data collected did not show statistically significant differences according to gender, but did show it according to age. Mardia’s multivariate normality tests indicate the absence of normality in the variables observed for both subsamples (see Table 1).
3.2 Second study
Corresponding to the second study, with n=586 participants, the extraction of factors by main axes was developed with the promax rotation method. Data from the correlation matrix were reflected as appropriate for this kind of study [Bartlett’s sphericity test (630) = 10920,894; p < 0.001; Kaiser-Meyer-Olkin Index = .931]. Five factors were obtained, according to the rule of eigen-values greater than one. The five-factor solution explained 49.22% of the variance. Regarding the content of the items, it can be seen that two factors were merged: “lack of emotional acceptance” and “lack of emotional awareness”, together with two more items that saturate this new factor. Given that the theoretical content of these items is similar to each other than compared to the other dimensions, it can be believed that for this reason the factorial process fails to distinguish them. Precisely to reject this argument, a factorial analysis again, but only with these items. All in all, the results showed again that these items corresponded to a single dimension or factor. Due to this, and due to the semantic characteristics of the items, the results show that it is more convenient to group them on a subscale. This is denoted in Table 2, which describes the data of the rotated configuration matrix; therefore, the first factor would be made up of fourteen items concerning the lack of emotional compression; the second would group eight items related to limited access to regulatory strategies; the third would consist of six items related to the difficulties in impulse control; the fourth would be grouped by four items, related to interferences in goal-directed behaviors; and the fifth would be grouped by four items related to the lack of emotional clarity.
Items | F1 | F2 | F3 | F4 | F5 |
Item 32 | .818 | ||||
Item 33 | .792 | ||||
Item 27 | .767 | ||||
Item 36 | .752 | ||||
Item 14 | .747 | ||||
Item 19 | .745 | ||||
Item 15 | .732 | ||||
Item 18 | .713 | ||||
Item 35 | .711 | ||||
Item 26 | .690 | ||||
Item 28 | .652 | ||||
Item 16 | .648 | .346 | |||
Item 13 | .539 | ||||
Item 17 | .509 | -.380 | .365 | ||
Item 30 | .831 | ||||
Item 29 | .578 | .825 | .388 | ||
Item 25 | .790 | ||||
Item 21 | .744 | -.335 | |||
Item 12 | .639 | ||||
Item 23 | .637 | -.334 | |||
Item 11 | .604 | ||||
Item 31 | .584 | ||||
Item 2 | .755 | ||||
Item 7 | .703 | .401 | |||
Item 8 | .700 | ||||
Item 6 | .688 | ||||
Item 1 | .652 | ||||
Item 10 | .499 | .314 | |||
Item 5 | .647 | ||||
Item 4 | .374 | .620 | |||
Item 9 | .585 | ||||
Item 3 | .395 | .525 | |||
Item 22 | .425 | .681 | |||
Item 24 | .332 | .586 | |||
Item 34 | .320 | .556 | |||
Item 20 | .508 | ||||
Variance of the rotated factor: | |||||
Factor | F1 | F2 | F3 | F4 | F5 |
4.49 | 3.47 | 2.12 | 1.71 | 1.21 | |
Factorial correlations: | |||||
Factor | F1 | F2 | F3 | F4 | F5 |
F1 | 1 | ||||
F2 | .66 | 1 | |||
F3 | -.07 | -.07 | 1 | ||
F4 | .58 | .57 | .02 | 1 | |
F5 | -.34 | -.37 | .42 | -.22 | 1 |
Note. Data in bold correspond to factor loads greater than 0.30 in absolute value; theoretically grouped factor correspondence.
3.3 Third study
With n=586 participants, this five-factor structure was examined using a confirmatory factor analysis, whose estimation procedure was performed by Unweighted Least Squares (ULS). Table 3 shows the final standardized values, whose adjustment was acceptable (Gaskin & Lim, 2016; Hu & Bentler, 1999) [x2(584) = 5512.724; GFI=.95; AGFI=.95; NFI=.94; RMR=.11], respectively. All the weights and factor correlations were statistically significant (p<.001), one by one.
F1: Lack of emotional understanding. | |
Item 32 | .760 |
Item 33 | .735 |
Item 27 | .784 |
Item 36 | .777 |
Item 14 | .707 |
Item 19 | .748 |
Item 15 | .710 |
Item 18 | .637 |
Item 35 | .786 |
Item 26 | .579 |
Item 28 | .713 |
Item 16 | .638 |
Item 13 | .492 |
Item 17 | .553 |
F2: Limited access to regulatory strategies. | |
Item 30 | .817 |
Item 29 | .735 |
Item 25 | .663 |
Item 21 | .694 |
Item 12 | .667 |
Item 23 | .505 |
Item 11 | .553 |
Item 31 | .665 |
F3: Impulse control difficulties. | |
Item 2 | .701 |
Item 7 | .771 |
Item 8 | .660 |
Item 6 | .626 |
Item 1 | .597 |
Item 10 | .502 |
F4: Interference in goal-directed behaviors. | |
Item 5 | .660 |
Item 4 | .687 |
Item 9 | .641 |
Item 3 | .662 |
F5: Lack of emotional clarity. | |
Item 22 | .424 |
Item 24 | .367 |
Item 34 | .686 |
Item 20 | .444 |
Note. The described composition of each item is found in the annex section of the research.
Finally, in terms of reliability, the internal consistency of each subscale was relatively good, since most of the Cronbach’s alpha coefficients and McDonald’s Omega coefficients showed higher values of .70 in most of the factors. For Factor 1 (fourteen items) α=.93, ω=.93; Factor 2 (eight items) α= .83, ω= .86; Factor 3 (six items) α=.80, ω=.81; factor 4 (four items) α=.75, ω=.76; and Factor 5 (four items) α=.60, ω=.61, which presented comparatively low reliability. The total reliability of the scale has a Cronbach’s alpha of .90 and a McDonald’s omega coefficient of .91, respectively.
4. Discussion
The purpose of the present study was to adapt DERS linguistically and culturally, as well as to evaluate the psychometric properties of the scale in the Ecuadorian university population. The results guarantee that the instrument has good psychometric properties of structure and internal consistency, being considered as good and consistent.
The new scale obtained is similar to the one originally conceived by the previous literature, since the Ecuadorian version subtly shows differences regarding dialectical compression, compared to the instrument originally recreated in the Anglo-Saxon language.
Due to the importance of developing a correct translation, for the conceptual equivalence of each item in the different cultural populations where the instruments are adapted (Carvajal et al., 2011), the scale is considered to measure the theoretical dimensions proposed by the authors, avoiding biases in the measurement of the construct. Precisely in the period of translation of the instrument, no significant problems in understanding were revealed, nor were unusual terminologies used. Precisely, the good conceptual and semantic equivalence in the new scale establishes the appropriation of the measure of difficulties in emotional regulation vis-à-vis the scientific community (Carvajal et al., 2011; Harkness & Schoua-Glusberg, 1998).
In addition to the above, the scale corresponds to an instrument of comfort and speed in its self-application. It is added the relevant aspect that was obtained through a pilot test, confirming the feasibility of the instrument for the Ecuadorian population. At this point it is necessary to mark the relevance that the scale uses a semantic structure that is easy to interpret, in order to avoid measurement errors (Argimon & Jiménez, 2013; Carvajal et al., 2011).
Regarding the factorial structure compared to the previously adapted Spanish and Latin scales, the results are similar to the Spanish adaptation (Hervás & Jódar, 2008), Chilean (Guzmán-González et al., 2014), and Argentinian (Medrano & Trógolo, 2014), where the five-factor solution is replicated due to the greater overlap of items grouped into the “lack of emotional compression” factor. However, the factor structure is different compared to the brief scale found in the Colombian con-text (Muñoz-Martínez et al., 2016). This result is influenced possibly by the quality of the responses obtained and the type of population to which the instrumental evaluation was submitted. Likewise, it is necessary to highlight that the factors present moderate and low correlations, which would be interpreted as a Relevant aspect of differentiated factors (Brown, 2006), observed in the same way in the Colombian adaptation of the instrument (Muñoz-Martínez et al., 2016); therefore, it solidifies theoretically what was mentioned by the original authors in the creation of the instrument, because the instrument presents a marked dimensionality of the con-struct depending on the context to which it is applied.
Regarding the limitations of the study, the possibility of bias in the study population is mentioned due to the lack of variability, which does not make possible a generalization in the entire population, since the study population was only made up of university students. Consequently, the sampling was non-probabilistic of se-lection by volunteers. Likewise, regarding the statistical part of the data analysis, despite the fact that there are currently much more robust and powerful factor extraction and rotation methods (Ferrando & Lorenzo-Seva, 2017), in the present study we opted to follow the same factorial design taken by the authors to replicate a similar internal structure. Another limitation is that the relationship with other variables could not be evaluated in terms of convergent and discriminant validity, because there are no Ecuadorian standardized instruments that measure the construct alternatively. It is suggested that for future studies the involvement of a diverse population, in order to know if the structure found in the study is replicated and corresponds to a universal determinant in the Ecuadorian population, as well as in the calculations from factor analysis, it is suggested that for future analyzes these should be discriminated by more robust and current computations of statistical stability, unlike those originally used by the authors.
As a conclusion, it is mentioned that, according to a translation and adaptation procedure of DERS, an instrument adapted to the Ecuadorian culture has been obtained in a university sample, whose psychometric values are established as adequate. Likewise, it is mentioned that in this version, the scale is similar to the original, in terms of structure and semantic compression, as well as the proper definition of the construct as ex-posed by the original authors.
Having an adapted scale in Ecuador, dealing with the difficulties of emotional regulation in an Ecuadorian university sample, responds to the impetuous need to have standardized instruments in the country that study this construct from a clear and simple language, due to the undeniable role of emotions on the academic particularities that the student is studying (González et al., 2009; Pekrun et al., 2000; Turner & Schallert, 2001), where emotional regulation plays an important role in academic performance (Diego et al., 2018; Medrano & Trógolo, 2014).