Social Anxiety Scale for Adolescents: Factorial invariance and latent mean differences across gender and age in Spanish adolescents

https://doi.org/10.1016/j.janxdis.2010.06.007Get rights and content

Abstract

Little is known about the factorial invariance across gender and age for self-report measures of social anxiety in adolescence. This study examined the factorial invariance and latent mean differences of the Social Anxiety Scale for Adolescents (SAS-A) across gender and age groups in 1570 Spanish adolescents (54% girls), ranging in age from 14 to 17 years. Equality of factor structures was compared using multi-group confirmatory factor analyses. Measurement invariance for the correlated three-factor model of the SAS-A was found across gender and age samples. Analyses of latent mean differences revealed that girls exhibited higher means than boys on two SAS-A subscales, Fear of Negative Evaluation and Social Avoidance and Distress-New (SAD-New). In addition, on the SAD-New subscale, the structured means significantly diminished from 14-year olds to 16- and 17-year olds and from 15-year olds to 17-year olds. Findings are discussed in terms of the use of the SAS-A with Spanish adolescents.

Introduction

Self-report questionnaires are a common method for assessing social anxiety in adolescence. The self-report method is especially important given the subjective and internalized nature of social anxiety (Kearney, 2005).

In recent years, increasing attention has been given to self-report measures of social anxiety in adolescence (e.g., García-López et al., 2008a, Storch et al., 2006, Viana et al., 2008, Zubeidat et al., 2007). Recent reviews conducted in Spain (García-López, Piqueras, Díaz-Castella, & Inglés, 2008) and the US (Kearney, 2005) have concluded that one of the most pertinent and empirically supported measures of social anxiety in adolescence is the Social Anxiety Scale for Adolescents (SAS-A; La Greca, 1999, La Greca and Lopez, 1998). In fact, the use of SAS-A in Spanish adolescents (Olivares et al., 2002) and American adolescents (Inderbitzen et al., 2004, La Greca and Harrison, 2005, Storch et al., 2004) is strongly recommended during screening or formal clinical evaluations (Kearney, 2005). This is because early recognition and intervention is necessary for adolescents with social anxiety to overcome this disorder and to prevent the development of other psychological disorders (Albano & Detweiler, 2001).

Given significant utility of the SAS-A for assessing adolescent social anxiety, the overall goals of the present study were to extend the existing evidence base for the SAS-A by examining the form and measurement invariance of the Spanish version of the SAS-A and the latent mean structure of the Spanish SAS-A across gender and adolescent age groups. Unless there is reasonable support for the measurement invariance of the factor structure of the SAS-A across gender and age groups, it may not be justified to compare responses of boys and girls or of different adolescent age groups (Bentler, 2005, Brown, 2006, Byrne, 2006).

According to Byrne (2006), there are five questions that must be addressed in order to evaluate an instrument's form invariance (questions 1, 2 and 3) and measurement invariance (questions 4 and 5). Specifically: (1) Is the factor structure of an instrument equivalent across populations? (2) Are certain paths in a specified instrument structure invariant across populations? (3) Does the factor structure of an instrument replicate across independent samples of the same population? (4) Do the items comprising a particular instrument operate equivalently across different populations (i.e., is the measurement model group invariant)? (5) And are the latent means of a particular construct in the model similar across populations? Dimitrov (2006) recommends testing form invariance of the measurement model across groups first, using the sample data for each group. If form invariance is observed, then measurement invariance is tested to make sure that scores on any construct have the same meaning for each of the compared groups.

The conceptual basis of the SAS-A, as well as the earlier Social Anxiety Scale for Children (La Greca, Dandes, Wick, Shaw, & Stone, 1988) originated from an adult model of social anxiety (Leary, 1983, Watson and Friend, 1969) that identified two components of social anxiety: social evaluative anxiety (also referred to as fear of negative evaluation), and social avoidance and distress. Factor analytic studies of children (e.g., Epkins, 2002, La Greca and Stone, 1993) and adolescents (e.g., La Greca & Lopez, 1998) provided further differentiation of the conceptual model of social anxiety as applied to youth. Three factors were identified: Fear of Negative Evaluation (FNE), consisting of eight items that assess fears, concerns, or worries regarding peer's negative evaluations (e.g., “I worry about what others say about me”); Social Avoidance and Distress in New Situations (SAD-New), consisting of six items that assess social avoidance and distress with new social situations or with unfamiliar peers (e.g., “I get nervous when I talk to peers I don’t know very well”); and Social Avoidance and Distress-General (SAD-General), consisting of four items that assess general social inhibition, distress, and discomfort (e.g., “I’m quiet when I’m with a group of people”).

From a conceptual perspective, the distinction between the SAD-New and the SAD-General subscales is an important one, as exemplified by distinctions in DSM-IV-TR diagnostic criteria (American Psychiatric Association, 2000). For example, children and adolescents with an avoidant disorder demonstrate social avoidance with unfamiliar persons (e.g., SAD-New), yet those with social phobia may evidence socially avoidant behaviors even with familiar persons (e.g., SAD-General). In terms of youths’ social experiences, social avoidance and inhibition in new situations is less problematic than more pervasive or generalized avoidance, as the latter may contribute to more restricted opportunities for normal socialization and friendships development (La Greca and Lopez, 1998, La Greca and Stone, 1993).

Since its initial development, the SAS-A has been translated into several languages. These translations include versions in Spanish (García-López, Olivares, Hidalgo, Beidel, & Turner, 2001), Portuguese (Cunha, Gouveia, Alegre, & Salvador, 2004), Turkish (Aydin & Sutcu, 2007), and Chinese (Zhou, Xu, Inglés, Hidalgo, & La Greca, 2007).

Accumulated evidence has shown that the three subscales of the SAS-A have adequate internal consistency (Cronbach's alpha greater than .70) and test–retest reliability. For example, La Greca and Lopez (1998) found internal consistencies ranging from .76 (SAD-General) to .91 (FNE); internal consistency for the total SAS-A score is also adequate (.87, in La Greca & Harrison, 2005). Furthermore, test–retest reliability estimates have ranged from .54 (SAD-General) to .78 (FNE) for a 2-month interval (La Greca, 1999) and from .55 (FNE) to .62 (SAD-General) for a 12-month interval (Storch et al., 2004).

Confirmatory factor analysis in a sample of American adolescents (La Greca & Lopez, 1998), provided initial evidence for the validity of the three-factor structure of the SAS-A. This factor model has been replicated in several samples of American adolescents (Inderbitzen et al., 2004, Inderbitzen and Walters, 2000, Myers et al., 2002, Storch et al., 2004), Turkish adolescents (Aydin & Sutcu, 2007), and Chinese adolescents (Zhou et al., 2008). In addition, Inderbitzen and Walters (2000) examined the form invariance of the SAS-A, finding that the factor loadings of the SAS-A items for the total sample were consistent across gender and grade. However, none of these studies have examined the measurement invariance of the SAS-A using multi-group confirmatory factor analysis in the framework of structural equation modelling (SEM).

Additional support for the validity of the SAS-A and its subscale scores comes from associations with other variables (i.e., convergent and discriminant validity, and test–criterion relationships). Specifically, SAS-A scores have been related to: (a) measures of social functioning such as, peer relations and acceptance, close friendships, and non-peer-related perceptions of support and competency (La Greca and Harrison, 2005, La Greca and Lopez, 1998); (b) general anxiety (Aydin and Sutcu, 2007, Inderbitzen and Walters, 2000) and dating anxiety (Glickman & La Greca, 2004); (c) depressive symptomatology (Inderbitzen and Walters, 2000, La Greca and Harrison, 2005); (c) other measures of social anxiety (Aydin and Sutcu, 2007, Cunha et al., 2004, Storch et al., 2004, Zhou et al., 2008); and (d) personality traits (Zhou et al., 2008). These studies also revealed differences in the strength of the correlations between the three SAS-A subscale scores and the other measures. For example, La Greca and Lopez (1998) found comparable associations between the three SAS-A subscales and adolescents’ social acceptance (r's = .41 to .48); however the SAS-A subscales were differentially related to the quality of adolescents’ close friendships, with SAD-General more strongly associated with poor friendship qualities than the FNE or SAD-New subscales. In a study by Inderbitzen and Walters (2000), adolescents’ scores on the FNE scale were more strongly related to their reports of general anxiety than were scores for SAD-New or SAD-General (for further details on the SAS-A subscales see Aydin and Sutcu, 2007, Cunha et al., 2004, García-López et al., 2001, Glickman and La Greca, 2004, Inderbitzen and Walters, 2000, La Greca and Harrison, 2005, La Greca and Lopez, 1998, Storch et al., 2004, Zhou et al., 2008).

The Spanish language version of the SAS-A also has received psychometric support across different adolescent samples. With respect to reliability, García-López et al. (2001) found that test–retest coefficients ranged from .75 (SAD-General) to .86 (total SAS-A) over 10 days, whereas Olivares et al. (2005) found internal consistency coefficients ranging from .80 (SAD-General) to .94 (FNE). Furthermore, the three-factor structure of the Spanish version of the SAS-A has been replicated in exploratory factor analysis (García-López et al., 2001) and single confirmatory factor analysis (Olivares et al., 2005).

With respect to validity of the Spanish SAS-A, García-López et al. (2001) examined associations with various measures of anxiety, avoidant personality disorder, and social anxiety, finding positive and statistically significant relationships between the SAS-A scores and other relevant measures. Furthermore, SAS-A total and subscales scores differentiated between boys and girls with and without social phobia, and among adolescents with circumscribed social phobia, generalized social phobia, and no diagnosis of social phobia (García-López et al., 2001). Finally, among Spanish adolescents who met criteria for generalized social anxiety, scores on the SAS-A were sensitive to treatment effects, with high effect sizes at post-treatment and 1-year follow-up (García-López, Olivares, & Hidalgo, 2005) as well as 5-year follow-up assessments (García-López et al., 2006).

Evidence from numerous studies indicates that adolescent girls report significantly more social anxiety than adolescent boys, especially with respect to fear of negative evaluation (Kearney, 2005). For example, García-López et al. (2001) found gender differences only on the FNE subscale score of the SAS-A, with Spanish girls reporting higher scores than boys. Other studies with adolescents found that girls report higher scores than boys on the SAS-A total and on the FNE and SAD-New subscales (Cunha et al., 2004, Inderbitzen and Walters, 2000, La Greca and Lopez, 1998, Myers et al., 2002, Olivares et al., 2005, Storch et al., 2004). Furthermore, these studies did not find gender differences on the SAD-General subscale.

Aside from gender differences, there may be age or grade differences in social anxiety, although previous findings are also inconsistent, even using samples of adolescents from the same country. Some studies have not found significant age or grade differences on any of the SAS-A scores (Cunha et al., 2004, La Greca and Lopez, 1998), whereas others found that younger adolescents reported higher scores than older adolescents on the FNE subscale and the total SAS-A score (Inderbitzen & Walters, 2000) or on the SAD-New subscale and the total SAS-A score (Storch et al., 2004). In Spain, García-López et al. (2001) did not find significant age differences on any of the SAS-A scores. However, Olivares et al. (2005) found that younger adolescents reported higher scores on the FNE subscale than older adolescents. Nevertheless, because the effect sizes for the age differences were very small, Olivares et al. (2005) cautioned that one cannot conclude that there are age differences in adolescents’ fear of negative evaluation.

Although prior research with the SAS-A has supported its factor structure and reliability, there are also a number of limitations of existing research that suggest several important gaps in the literature. The present study addresses several of these gaps.

First, prior research on the SAS-A has not examined the normality or distribution of the SAS-A item scores. Understanding normality of the SAS-A item scores is important because non-normality can greatly affect the results of confirmatory factor analysis (CFA) that is based on maximum likelihood (ML) estimation method (Finney & DiStefano, 2006). Therefore, it is important to recognize the assumptions associated with the estimation method, demonstrate the conditions under which results are robust to violations of these assumptions, and specify the procedures that should be employed when assumptions are not met. Specifically, Finney and DiStefano (2006) recommend that investigators assess the distribution of the observed variables (e.g., the SAS-A item scores) prior to conducting CFA in order to make an informed decision about the best estimation method. Thus, in the present study, the distribution of the SAS-A item scores was evaluated.

Second, previous research has not examined whether the obtained three-factor model of social anxiety, as measured with the SAS-A, is invariant across gender and age groups. Rather, previous research that focused on multi-group comparisons (e.g., gender or age differences) of SAS-A scores typically assumed that this instrument operates in exactly the same way across the groups of interest (e.g., boys and girls) and that the underlying construct (i.e., social anxiety) has the same theoretical structure and psychological meaning across the groups of interest. Thus, to date, information is lacking on whether the SAS-A measures the same components of social anxiety with equal validity for boys and girls and for adolescents of different ages.

In fact, some researchers have argued that establishing measurement invariance is necessary for a between-group difference to be unambiguously interpreted. This is because the between-group difference may reflect either a “true” difference on the construct of interest or different psychometric properties of the scale items (Brown, 2006). Such concerns about factor structure invariance are most appropriately analyzed using multi-group confirmatory factor analysis (MGCFA). MGCFA is an extension of single CFA that can determine whether (and how) the factor structure of the SAS-A varies according to gender and age group (Brown, 2006, Byrne, 2006). Therefore, the present study used MGCFA to examine whether the components of the SAS-A are operating equivalently across adolescent age and gender and also to analyze the latent mean differences in SAS-A scores across gender and age groups.

In summary, the present study focused on the Spanish version of the SAS-A, using a sample of Spanish adolescents. The key study goals were: (a) to examine the internal consistencies of the SAS-A total and subscale scores; (b) to examine the normality or distribution of the SAS-A items by gender and adolescent age groups; (c) to extend evidence for the validity of the SAS-A's internal structure by analyzing form and measurement invariance; and (d) to analyze the invariance of latent mean structures on the SAS-A across gender and adolescent age groups by means of multi-group CFA.

Based on the goals described above, the following hypotheses were examined: (a) that the total and subscale scores on the Spanish version of the SAS-A will demonstrate adequate internal consistency (e.g., Cronbach's alpha above .70); (b) that the three-factor correlated structure of the SAS-A will be equivalent across gender and adolescent age groups; and (c) that there will be latent mean differences on the SAS-A subscales across gender and age groups, although the exact pattern of the latent mean differences is an open research question. Previous studies examined gender and age or grade differences based on observed mean scores for the SAS-A, but not on latent mean scores. Also, as noted previously, findings across studies have been inconsistent with respect to gender and age or grades differences in SAS-A scores.

Section snippets

Participants

The research study took place in two provinces in southeast Spain during the 2006–2007 academic year. The province of Alicante served 71,565 students in 241 schools and the province of Murcia served 65,264 students in 224 schools (Spanish Ministry of Education and Science, 2007).

Cluster random sampling was performed throughout the five geographical areas of the provinces of Alicante and Murcia, Spain: center, north, south, east and west. Seventeen high schools from rural and urban areas, 14

Internal consistency

Coefficient alpha reliabilities for the SAS-A total and the three subscales were calculated in all subsamples (i.e., gender and age groups). These coefficients ranged from .71 (SAD-New) to .89 (SAS-A total) for boys, and from .76 (SAD-New) to .90 (SAS-A total) for girls. Furthermore, the range of coefficients by age groups was as follows: .72 (SAD-New) to .89 (SAS-A total) for 14-year olds; .71 (SAD-New) to .90 (SAS-A total) for 15-year olds; .76 (SAD-New) to .90 (SAS-A total) for 16-year olds;

Discussion

Social anxiety disorder is a common mental health disorder that affects children, adolescents, and adults. Although increasing attention has been given to the assessment of social anxiety in adolescence (see García-López et al., 2008b, Kearney, 2005) very little is known about cross-gender and cross-age equivalence in the expression of social anxiety among adolescents. This gap in knowledge is of great concern because social anxiety disorder is one of the most common emotional disorders among

Acknowledgements

This research was supported by a grant from the Spanish Ministry of Science and Innovation (SEJ2004-07311/EDUC) to the first author. A special thank to Timothy A. Brown, Professor of Psychology at Boston University, for his helpful comments on the article.

References (56)

  • P.M. Bentler

    EQS 6.1: Structural Equations Program Manual

    (2005)
  • T. Brown

    CFA with equality constraints, multiple groups, and mean structures

  • B.M. Byrne

    Structural equation modeling with EQS: basic concepts, applications and programming

    (2006)
  • G.W. Cheung et al.

    Evaluating goodness-of-fit indexes for testing measurement invariance

    Structural Equation Modeling

    (2002)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • M. Cunha et al.

    Assessment of social anxiety in adolescence: the Portuguese version of the SAS-A

    Psychologica

    (2004)
  • P.J. Curran et al.

    The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis

    Psychological Methods

    (1996)
  • V.D. Dimitrov

    Comparing groups of latent variables: a structural equation modeling approach

    Work: a Journal of Prevention, Assessment & Rehabilitation

    (2006)
  • C.C. Epkins

    A comparison of two self-report measures of children's social anxiety in clinic and community samples

    Journal of Clinical Child and, Adolescent Psychology

    (2002)
  • S.J. Finney et al.

    Non-normal and categorical data in structural equation modeling

  • L.J. García-López et al.

    Brief form of the Social Phobia and Anxiety Inventory (SPAI-B) for adolescents

    European Journal of Psychological Assessment

    (2008)
  • L.J. García-López et al.

    A pilot study on sensitivity of outcome measures for treatments of generalized social phobia in Spanish adolescents

    International Journal of Clinical and Health Psychology

    (2005)
  • L.J. García-López et al.

    Psychometric properties of the Social Phobia and Anxiety Inventory, the Social Anxiety Scale for Adolescents, the Fear of Negative Evaluation Scale and the Social Avoidance and Distress Scale in an adolescent Spanish-speaking sample

    Journal of Psychopathology and Behavoiral Assessment

    (2001)
  • L.J. García-López et al.

    Trastorno de ansiedad social en la infancia y adolescencia: Estado actual, avances recientes y líneas futuras [Social anxiety disorder in childhood and adolescence: current state, recent advances, and future lines]

    Behavioral Psychology/Psicología Conductual

    (2008)
  • A.R. Glickman et al.

    The Dating Anxiety Scale for Adolescents: scale development and associations with adolescent functioning

    Journal of Clinical Child and, Adolescent Psychology

    (2004)
  • M. Gren-Landell et al.

    Social phobia in Swedish adolescents: prevalence and gender differences

    Social Psychiatry and Psychiatric Epidemiology

    (2009)
  • G.R. Hancock

    Structural equation modelling methods of hypothesis testing of latent variable means

    Measurement and Evaluation in Counseling and Development

    (1997)
  • C. Hayward et al.

    The developmental psychology of social anxiety in adolescents

    Depression and Anxiety

    (2008)
  • Cited by (0)

    View full text