Participants and Procedure

Data was collected from a sample of 840 adolescents between ages 12 and 17 years of age, with a mean age of 14.3 (SD=1.27).Of them, 47% were male and 53% were female. Instruments were applied in distinct educational institutions within the Guadalajara Metropolitan Area, prior informed consent from the Institutions’ Principals.

For statistical analyses, the SPSS v.21. Statistical Program was used. Initially, an exploratory factor analysis was carried out and the items internal consistency was analyzed as well. Subsequently, the factor analysis was confirmed with AMOS Program and correlation analyses were undertaken to evaluate the scale’s convergent and divergent validity.

Instruments

The following instruments were applied in addition to a questionnaire to identify participants’ sociodemographic characteristics.

READ, Resilience Scale for Adolescents (^{Hjmedal, et al., 2006}). The 28-item original version of the READ is composed of positively phrased items, with a 5-point Likert-type structure, where 1 means Totally Disagree and 5 Totally Agree. Higher scores on the READ indicate higher level of resiliency (^{Hjemdal et al., 2006}).

The original scale is composed of 5 factors: Personal Competence, Social Competence, Structured Style, Family Cohesion and Social Resources. It maintained the conceptual content from its predecessor to evaluate resilience in adults (the Resilience Scale for Adults of ^{Friborg, Barlaug, Martinussen, Rosenvinge, & Hjemdal, 2005}).

The Personal Competence factor measures the individual levels of self-esteem, self- eficacy, self-acceptance, hope, determination, realistic life-orientation, and ability to follow daily routines, same as to plan and organize. Social Competence makes reference to extraversion, social skills, good humor, capability on starting conversations and flexibility on social environments. The Structured Style factor measures the level of preference in which an individual plans and structures his daily routines. The Family Cohesion measures the level in which values are shared; there is also social support and the ability of the family to maintain a positive perspective. Finally, the Social Resources factor evaluates perception on access and availability of external support, such as friends (^{Hjemdal et al., 2006}).

The authors of the scale reported acceptable reliability indices for the ive factors assessed in the READ; the indices are: Personal Competence (α=.85), Social Competence (α=.82), Structured Style (α=.69), Family Cohesion (α=.85), and Social Resources (α=.78).

EQ-i: YV, Bar-On Emotional Quotient Inventory, Youth Version (^{Bar-On & Parker, 2000}). The Spanish version was developed and validated by ^{Ferrandiz, Hernandez, Bermejo, Ferrando, and Sainz (2012)}; its purpose is to measure social and emotional competences in children and adolescents aged between 7 and 20 years.

It is a 60-item self-report instrument in a 4-point Likert scale, whose answer vary from 1 (It never happens to me) to 4 (It always happens to me) and evaluates 5 factors and each in turn include several subscales: intrapersonal competencies (emotional self-awareness, assertiveness, self-consideration, self-fulfillment and emotional independence), interpersonal competencies (empathy, social responsibility and interpersonal relationship), stress management (stress tolerance and impulse control), adaptability (problem-solving, emotional validity and flexibility) and mood (joy and optimism) (Bar-On, 1997; ^{Ferrandiz et al., 2012}).

Reliability indices found in this study are α =.665 for intrapersonal competencies, α = .822 for interpersonal competencies, α = .627 for stress management, α = .827 for adaptability and α = .872 for mood.

RSES, Rosenberg Self-Esteem Scale (^{Rosenberg, 1965}). Translation into Spanish by ^{Martin-Albo, Nunez, Navarro, and Grijalvo (2007)}. It is a 10-item response Likert-type scale, whose range varies between 1 (Strongly Disagree) and 4 (Strongly Agree). It is a scale that even when it was conceived as bifactorial, it shows better adjustment results when it is considered as unifactorial. In this study, internal reliability indices of α = .799 were found.

CMAS-R, Children’s Manifest Anxiety Scale-Revised (^{Reynolds & Richmond, 1997}). It is an instrument that evaluates the level and nature of anxiety in children between 6 and 19 years of age.

It consists of 37 items included in three subscales: Physiological Anxiety (associated to physiological anxiety symptoms such as sleeping problems, nauseas or fatigue), Worry and Hypersensitivity (related to worrying compulsively about several things; most of them are rather vague and are not well defined in a child’s mind, in addition to fear of being hurt or isolated in an emotional way), and the Social Worries Subscale (it refers to distracting thoughts, as well as fears of social and interpersonal nature that lead to concentration and attention disorders). Additionally, it counts with a Lie Scale designed to detect approval, social convenience or deliberated falsification of answers (Reynolds & Richmond, 1985). A reliability index of α = .789 was found (Cronbach’s Alpha) in this study.

CDI Children’s Depression Inventory (^{Kovacs, 2004}). It is a 27-item self-report instrument that evaluates depressive symptoms in individuals between 7 to 16 years of age. It consists of two scales: Dysphoria, which considers elements such as depressive mood, sadness, worry and the Negative Self-Esteem Subscale, that includes inefficiency judgments, ugliness or malice (^{Kovacs, 2004}). An index of α = .806 (Cronbach’s Alpha) was obtained in this study.

Exploratory Factor Analysis

An Exploratory Factor Analysis of the READ structure was conducted through the main components method, and based on eigenvalues greater than 1. The Kaiser-Meyer-Olkin (KMO) showed an adequacy index of 0.92. Bartlett’s test of sphericity was significant, to the level of p<.01 = 6792.92 gl = 378), demonstrating that intercorrelations between items was appropriate.

Conducting the analysis resulted in a 7 components matrix, in which items 4 (“I’m satisfied with my life so far”) and 12 (“I stop caring about things when they seem impossible to change”) were part of a factor that would correspond to personal competencies in a theoretic model; however, given the nature of both, item 4, which is linked to a vital satisfaction, as well as item 12, whose syntactic structure could have led participants to interpret it as a negative response, it was decide it to eliminate them from the scale.

Table 1 shows the matrix of 7 resultant components after said items were eliminated. It should be noted that items 1, 2 and 7, which make reference to goal-orientation, were assigned independently from personal competencies, where they belong in the model of the authors of the scale. For this reason, it was decided to integrate them as another likely factor within the scale’s structure.

In general, factor analysis data evidence that most of the items are properly associated to the theoretical model’s dimensions.

The factor related to Family Cohesion was composed of the 6 items from the original scale, to which item 18 (“We have rules that simplify daily life in our family”) that refers to family was added and, even when it belonged to the Structured Style, could have been interpreted based on the support provided by them. Additionally, we should note that despite item 21 showed a slightly major factor load in the Personal Competency Factor, it shows a proper factoring in the Family Cohesion component to which this item corresponds in the original scale.

Each of the Social Competence and Social Resources factors were composed by 5 items of the original version, with saturations fluctuating between .43 and .79.

The factor corresponding to Personal Competences was the one it showed major changes, given that only 4 of the 8 items that compose it in the original scale, could group together; at the same time, other 3 grouped together independently, having as a common factor aspects related to achieving goals, for which it was considered as a new factor that can be applied to the version validated in Mexico.

Finally, the factor related to the structured style presented a group composed of only 2 items, given the item 2 was sorted into the new factor (goal-orientation), and item 18 was sorted into to the Family Cohesion; because of that, we decided to eliminate this factor of the scale, also this factor was the weakest in the original version (^{Hjemdal et al., 2006}; ^{Von Soest et al., 2010}), and probably the nature of the items could be focus on exploring elements associated with personality.

Confirmatory Factor Analysis

The confirmatory factor analysis was conducted by means of the statistical package AMOS in a combined model with the interrelation of the 5 factors found in the Exploratory Factor Analysis.

After the analysis, and with the intention to reduce the number of items, it was decided to eliminate the items that showed the lowest factorial standardized saturation. Item 18 of the Family Cohesion component and item 9 in the Social Resources component were eliminated. Eliminating those two items improved the reliability indices in both scales, resulting in a Cronbach’s Alpha of .842 for Family Cohesion and .710 for Social Resources.

In the Figure 1 is the final structure of Confirmatory factor Analysis with 22 items version. Adjustment indices were adapted ( = 558.762 gl = 199; CFI = .932; NFI = .899; IFI = .933; TLI = .921; RMSEA =.046).

Figure 1. Conirmatory Factor Analysis of the READ 

Convergent and Divergent Validity

A correlation analysis was conducted with other psychological and social variables, including those variables of a sociodemographic type, this with the purpose to carry out the convergent and divergent validity of the scale.

Convergent validity was carried out correlating subscales of the final version of READ with measures about socioemotional competencies (EQi-YV) and Self-Esteem (RSES). Results are shown in Table 2.

Divergent validity was carried out by correlating factors of the READ with anxiety measures (C-MAS) and Depression (CDI). Results are shown in Table 3.

Finally, a variance analysis (ANOVA) among measures was conducted with the purpose to identify the possible existence of gender differences among factors that compose the scale; results were that only statistically signiicant differences were found in the factor related to personal competence F (1,837) = 4.541 p<.05 where males showed higher scores, as well as in the factor corresponding to social resources F (1,837) = 5.796 p<.05 where females scored higher.

Discussion

The interest in resilience has been largely motivated by the possibility to understand and identify protective factors and mechanisms that are essential to positive adaptation and that could aid in the prevention of psychological problems (^{Hjemdal et al., 2007}; ^{Masten, 2001}; ^{Laird, 2004}; ^{Waaktaar & Torgensen, 2010}). Particularly publications related to this topic have substantially increased in about 85% when compared to those in the last decade (^{Jowkar, Friborg & Hjemdal, 2011}; ^{Hjemdal et al. 2007}; ^{Windle et al., 2011}).

In this study, the Spanish version of the READ was assessed and the scales were modified by taking into account the results from factor analyses. A 26-item version of the READ with proper factor structure, psychometric properties and acceptable validity is proposed as a valid measure to evaluate protective factors related that develop to resilience in Mexican adolescents’ ages 12 to 17 years-old. This is an important advance in resilience research as most of the research has been conducted with mixed and/or primarily Caucasian samples (^{Laird, 2004}).

The current study supported a ive-factor solution for the READ. It is important to emphasize the presence of an additional factor to the original scale in Mexican population. This factor is related to goal-orientation and appears to be disassociated, in terms of factor, to the Personal Competence subscale with a possible weakening of this scale.

The manifestation of this factor is relevant because coincides with theoretical contributions from other resilience models that propose not only the importance of visualizing the future and having goals as essential to discovering the meaning of life (^{Vanistendael, 2005}), but the perception of our capabilities to achieve them (^{Saavedra & Villalta, 2008}; ^{Theis, 2003}; ^{Vanistendael, 2003}).

In the current study, results from conirmatory factor analyses revealed that the factor structure of the original 28-item measure could beneit from modiication. The performance of item 18, associated to family rules that simplify everyday life and that originally belonged to the Structured Style scale, was is associated to the Family Cohesion factor, coinciding with previous validations of the scale in Norwegian and Italian population (^{Von Soest et al., 2010}; ^{Stratta et al., 2012}).

Due to poor it indices, the factor that required major changes was Personal Competences. This also happened in the study of ^{Von Soest et al. (2010)} However, in their study it was only required the elimination of one item, whereas in the present study three of the eight items need to be loaded into the new factor of Orientation to goals.

Interesting was the fact that ^{Von Soest et al. (2010)} and ^{Stratta et al. (2012)} also reported problems in some of the items that were changed or removed in the current study. For example, item # 4 “I am satisfied with my life until now” was removed from both studies. Item # 1”I reach my goals if I work hard”, was relocated to the Orientation to goals factor but ^{Von Soest et al. (2010)} decided to eliminate it from the scale. Item #12 “When it is impossible for me to change certain things I stop worrying about them” also showed poor it indices as in the ^{Stratta et al. (2012)} study and was deleted from both studies.

A widely used method for evaluating construct validity in speciic samples is to examine both, positive and negative correlations with other established scales. In this study, divergent validity was supported as the READ was significantly negative correlated with anxiety symptoms (CMAS-R) and especially with depressive symptoms (CDI). These findings have also been reported by other studies (^{Hjemdal et. al 2006}; ^{Hjemdal et al., 2007}; ^{Von Soest et al., 2010}). A negative correlation between anxiety, depressive symptoms, and resilience was expected as it is well known that experiencing a negative mood, negative future outlook, physiological symptoms of anxiety and/or extreme worry, among other symptoms could decreased an individuals´ level of resilience (^{Dozois & Westra, 2004}; ^{Gordon, 1987}; ^{Kendall & Suveg, 2006}). Furthermore, there is evidence that supports the predictive value of the READ for depressive symptoms in young adolescents. Results from the study by ^{Hjemdal et al. (2007)} showed that adolescents who reported higher levels of resilience exhibited significantly lower levels of depressive symptoms, even when controlling for gender, age, the number of stressful life events including being bullying, and levels of social anxiety.

The predictive value of the READ for depressive symptoms is of important value, especially for the population of Mexican adolescents. A study by ^{Benjet, Borges, Medina-Mora, Zambrano, & Aguilar-Gaxiola (2009)} showed that almost 40% of Mexican adolescents reported a 12-month disorder, being anxiety and depression two of the most common. Depression also appears to be one of the top mental health problems reported by Mexican adults (^{Medina-Mora et al., 2003}).

A particular strength of this study was that it also used scales to evaluate convergent validity, which were expected to correlate positively with the READ. To our knowledge, no published studies on the READ have incorporate this besides the study of ^{Von Soest et al. (2010)} that included a 10-item short version of the Parental Bonding Instrument. As expected, results from this study showed statistically signiicant correlations between participants´ self-esteem, social and emotional competencies and their level of resilience. Actually, self-esteem was one of the factors included in the resilience scaled developed by ^{Gonzalez (2011)}.

Regarding gender differences, males reported statistically higher levels of personal competence than females; and females reported statistically higher levels of social resources than males. These differences were also found in the studies by ^{Hjemdal et al. (2006)} and ^{Von Soest et al. (2010}). Similarly, ^{Stratta et al. (2012)} found that males scored signiicantly higher in the Personal Competence factor but they also found that males reported higher total scores. It has been reported in previous literature that men perceived themselves as more competent and women are generally using more social support and resources than men (^{Werner, 1989}), and Mexican culture in some ways also support this.

Limitations and Further Research

The current study could have benefit from a more representative sample including Mexican adolescents of a wider age range, and from including adolescents from all regions of the country including rural and urban areas. Further research should also consider the acculturation process as well as the social and political factors that might contribute to better understand resilience in the Mexican culture (^{González et al., 2008}).

As suggested by other researchers, further studies should also tried to ind a sample of “truly resilient adolescents” who have experienced important stressors (e.g., one’s own illness or an illness in the family) but that still have managed to cope (^{Hjemdal et al., 2006}; ^{Von Soest et al., 2010}). This will be particularly interesting and will allow differentiating speciic characteristics and levels of resilience for different groups of Mexicans.

It will be important as well that further research include the evaluation of at least one item of the personal competence factor and one of the factor related to goal orientation, in order to explore the internal validity and reliability of the instrument.

Finally, further studies should also include indicators of SES and school grades in order to evaluate if there are any signiicant relationships such as the ones found in the study by ^{Von Soest et al. (2010)}. Indicators of SES can also be explored through more comprehensive questions related to the new factor of goal-orientation. It will also be ideal to include estimates of test-retest reliability in order to give additional insight into the scale´s reliability.