Anger expression in Swiss adolescents: Establishing measurement invariance across gender in the AX scales

Anger expression in Swiss adolescents: Establishing measurement invariance across gender in the AX scales

Journal of Adolescence 35 (2012) 1013–1022 Contents lists available at SciVerse ScienceDirect Journal of Adolescence journal homepage: www.elsevier...

216KB Sizes 0 Downloads 48 Views

Journal of Adolescence 35 (2012) 1013–1022

Contents lists available at SciVerse ScienceDirect

Journal of Adolescence journal homepage: www.elsevier.com/locate/jado

Anger expression in Swiss adolescents: Establishing measurement invariance across gender in the AX scales Daniel Zimprich*, Anna Mascherek Institute of Psychogerontology, Methods and Prevention Research, University of Erlangen-Nuremberg, Naegelsbachstrasse 25, D-91052 Erlangen, Germany

a b s t r a c t Keywords: Anger expression Measurement invariance Gender differences Adolescence

The present study examined measurement invariance in the three anger expression subscales of the STAXI (Spielberger, 1988) with respect to gender. In a sample of 576 male and 531 female students, strict measurement invariance was found. For all three anger expression factors, no differences in variances or factor correlations were found. A large negative relation between Anger-Out and Anger-Control emerged. Girls reported significantly lower levels in Anger-Out and Anger-Control than boys. Results suggest that the questionnaire functions the same way across gender. Also, boys and girls exhibited the same range of interindividual differences in anger expression. The negative relation between Anger-Out and Anger-Control suggests that high levels of Anger-Out might be an obstacle in controlling anger. Lower levels of Anger-Out and -Control in girls suggest that girls might need less control because they express anger less outwardly. Ó 2012 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.

Introduction Anger belongs to the most frequent everyday emotions that individuals report (e.g., Schimmack, 2002). Compared to other emotions (e.g., joy, love), anger may be characterized as an emotion that is experienced as arousing, unpleasant, and aversive (cf. Averill, 1982, 1983; Berkowitz, 1999). Due to its aversive nature, anger represents an emotional state that has to be dealt with in one way or another. As several authors have argued, anger may be expressed either inwardly or outwardly (Averill, 1982; Funkenstein, King, & Drolette, 1954; Spielberger, 1988). Anger-In indicates an active suppression of angry feelings by the individual. Indicators of internally expressed anger are, for example, holding things in, keeping grudges, and concealing anger. By contrast, Anger-Out refers to the tendency to respond with either physical or verbal aggression when angry. Examples of externally expressed anger are slamming doors, saying nasty things, and telling off others. Spielberger (1988, 1996; Spielberger et al., 1985) distinguished an additional aspect of anger expression – Anger-Control. This aspect reflects attempts to manage the expression of anger by, for example, controlling one’s temper and behavior. The importance of different forms of expressing or controlling anger stems from the fact that the way individuals manage their angry feelings is related to health outcomes, e.g., blood pressure (Bongard & al’Absi, 2003), migraine headaches (Gerhards, 1992), cancer (McKenna, Zevon, Corn, & Rounds, 1999), or eating disorders (Fassino, Abbate-Daga, Pierò, Leombruni, & Rovera, 2003). Moreover, the way anger is expressed is associated with aggressive behavior (Furlong & Smith, 1994) and violence (Eckhardt, Jamison, & Watts, 2002). With respect to adolescents, Musante and Treiber (2000), for example, reported that adolescents inwardly expressing anger drank alcohol more frequently, spent fewer hours per week in aerobic activity, and were less physically active than their peers contrarily. Eftekhari, Turner, and Larimer (2004) demonstrated that in incarcerated * Corresponding author. Tel: þ49 (0) 9131 85 26526. E-mail address: [email protected] (D. Zimprich). 0140-1971/$ – see front matter Ó 2012 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.adolescence.2012.02.008

1014

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

adolescents outwardly expressing anger was significantly associated with both alcohol and marijuana use. Apart from substance abuse, the way of expressing anger may also have psychological consequences in adolescents. Lutwak, Panish, Ferrari, and Razzino (2001) found that in 265 students shame-proneness was positively related to expressions of inward anger, whereas guilt-proneness was negatively related to outward anger, but positively related to anger control. Cautin, Overholser, and Goetz (2001) reported that adolescents who internalized their anger were more likely to be depressed and to experience feelings of hopelessness. Also, adolescents who internalized their anger made more serious suicide attempts than did those who externalized their anger. Of course, the way anger is expressed also has consequences on social relationships (cf. Hess & Kirouac, 2000). In order to psychometrically assess individual differences in anger expression and control, Spielberger et al. (1985) developed the Anger Expression Inventory (AX), a self-report questionnaire which is designated to measure Anger-In, Anger-Out, and AngerControl. Subsequently, the Anger Expression Inventory became part of the State-Trait Anger Expression Inventory (STAXI), which, in addition to anger expression and control, measures state anger and trait anger, which comprises angry temperament and angry reaction (Spielberger, 1988, 1996). Accordingly, the STAXI consists of six subscales that were designed to map on six correlated factors. To date, only few studies examined the factorial structure of the STAXI in adolescents (e.g., Armstead & Clark, 2002; del Barrio, Aluja, & Spielberger, 2004; Forgays, Forgays, & Spielberger, 1997; Fuqua et al., 1991; Musante, Treiber, Davis, Waller, & Thompson, 1999). Fuqua et al. (1991), for example, reported that in a sample of 455 college students, results of several exploratory factor analyses were similar to the factorial structure postulated for the STAXI. Forgays et al. (1997) established the six proposed factors in a sample of 714 university students. However, they also found evidence for an additional seventh factor, mainly composed of state anger items. By contrast, Armstead and Clark (2002) failed to replicate the factor structure of the anger expression subscales, although their sample of 86 African American adolescents was comparatively small. Consistent with other studies, though, Anger-Control items loaded on a common factor (e.g., Musante et al., 1999). Recently, del Barrio et al. (2004) described the psychometric properties of a new psychometric measurement instrument of anger in children and adolescents (STAXI-CA), which was derived from the original STAXI. Using both exploratory and confirmatory factor analysis, del Barrio et al. (2004) reported a three factor structure of the anger expression (Anger-In, Anger-Out) and Anger-Control items for both male and female children and adolescents. To summarize, the available evidence seems to largely support the proposed factorial structure of the STAXI. Sex differences in anger expression Conventional wisdom suggests that women are more “emotional” than men (cf. Brody, 1997; Sharkin, 1993). In line with this lay assumption, a number of studies have found women to be more expressive of most positive and negative emotions than men (e.g., Brody & Hall, 2000; Fischer & Manstead, 2000; Kring, 2000; Kring & Gordon, 1998). Fairly well-established theories assert that boys and girls learn different rules for the expression of emotion, but not necessarily for the experience of emotion (Brody, 1985). Moreover, sex roles might contribute to sex differences in emotional expressivity and in the interpretation of people’s emotional expressions (e.g., Brody & Hall, 2000; Plant, Kling, & Smith, 2004). Besides, expressive behavior in social situations is believed to be influenced by socially and culturally determined display rules d social and cultural standards, which might be different for men and women, about how and when to express emotion (Ekman, 1992). Whether the general notion of women being more emotionally expressive also holds for anger seems questionable (see Kring, 2000). In fact, a common assumption is that women, as compared to men, might have more difficulties in expressing anger (Sharkin, 1993). Plant, Hyde, Keltner, and Devine (2000), for example, reported that participants in their studies believed that men expressed and experienced anger more often than women did. Different socialization is usually identified as a key factor for explaining these sex-related differences in anger expression (cf. Brody, 1997). Although women are socialized to show their emotions more openly than are men, women’s open expression of anger is viewed as “unfeminine.” By contrast, expression of anger by men is considered to be male-like. However, the assumption of men expressing more anger than women is not consistently supported by empirical data, and has only been shown in a few studies (e.g., Deffenbacher, Oetting, Lynch & Morris, 1996). Moreover, while it might be that men express more anger through vocal, facial, and behavioral modalities than do women, the sex difference is much less pronounced regarding the frequency and intensity of anger expression in studies using self-report questionnaires (Brody & Hall, 2000). One reason for this result is seen in the fact that questionnaire items typically leave open the social context in which anger occurs (Brody, 1997), which might play a vital role due to the interpersonal consequences of expressing anger (Averill, 1983). Whereas women are more anxious about the possible negative consequences for others when expressing anger (Eagly & Steffen, 1986), and are more likely to empathize with the victim (Timmers, Fischer, & Manstead, 1998), men tend to not anticipate such negative reactions, and may even expect positive outcomes of their anger expression, such as admiration (Campbell & Muncer, 1987). In line with the role of social context, once situational information is provided, women express anger to a much lesser extent than men (Evers, Fischer, Rodriguez Mosquera, & Manstead, 2005). With respect to the role of gender in anger expression as measured by the AX subscales of the STAXI, previous research has led to inconsistent findings. Some researchers found that males and females did not differ on the STAXI dimensions of AngerOut, Anger-In, and Anger-Control (e.g., Armstead & Clark, 2002; Spielberger, 1988, 1996; Stoner & Spencer, 1987). By contrast, other researchers reported sex differences in different aspects of anger expression (e.g., Fisher et al., 1993; Schwenkmezger, Hodapp, & Spielberger, 1992; Spielberger et al., 1985; Suter, Byrne, & Byrne, Howells, & Day, 2002). For example, in a large sample of high school students, Spielberger et al. (1985) found that males scored higher on inwardly expressed anger than did females. However, they found no sex differences for outwardly expressed anger. In different validation samples for the German version of the STAXI, Schwenkmezger et al. (1992) found sex differences in anger expression. These results did not

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

1015

show a consistent sex-pattern across samples, however. Recently, Suter, Byrne, Byrne, Howells, and Day (2002) reported sex differences in angry prisoners, where women displayed higher scores on Anger-In, and Anger-Out, while men scored significantly higher on Anger-Control. To summarize, previous studies have led to inconclusive results. While in some studies evidence has been reported that women have more difficulties in acknowledging and expressing anger than man, in other studies no sex differences have been found. As noted above, one reason for this mixture of results might be that self-report questionnaire items do not provide information about the social context of expressing anger. In the present study we consider another possible reason for “missing” sex differences in anger expression: A comparison of anger expression, as measured with the STAXI, in men and women relies on the assumption that the AX subscales have the property of measurement invariance (MI) across gender (cf. Meredith & Horn, 2001). As Horn and McArdle (1992, p. 117) have defined it, MI refers to “whether or not, under different conditions of observing and studying phenomena, measurement operations yield measures of the same attribute.” Hence, the property of MI ensures that a measure is unbiased across groups of individuals classified by a selection variable, e.g., sex. More specifically, sex differences in anger expression (or their absence) could be due to differential item functioning in women and men. Where could such differential item functioning stem from? One possibility would draw on response styles (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Social desirability, for example, was defined by Ganster, Hennessey, and Luthans (1983) as the “tendency for an individual to present him- or herself, in test-taking situations, in a way that makes the person look positive with regard to culturally derived norms and standards” (p. 132). Because the purpose of this response style is to intentionally create a positive impression of the individual by inflating responses relative to the true score, social desirability can be described as “faking good.” Now if women and men answer self-report items in a manner that is influenced not only by their actual way of expressing anger but also by differential amounts of social desirability, this would lead to items that function differentially in women in men. Previous studies have found that there may be sex differences in social desirability (e.g., Crandall, Crandall, & Katkovsky, 1965; Davies, 2001). Thus, it is possible that social desirability biases lead men and women to endorse particular gender-relevant traits (e.g., anger) differentially. The total scores (or factors) of anger expression would then contain a mixture of both the construct in question and a method effect (e.g., social desirability). In line with this assumption, previous studies have reported that participants show social desirability in answering the AX scales (McEwan, Davis, MacKenzie, & Mullen, 2009). Without MI being tested for, one can not disentangle whether group differences reflect real differences between the groups in the respective construct or whether measurement properties differ systematically between groups due to, for example, social desirability. MI is an issue of degree, which, borrowing from Meredith’s (1993) terminology, ranges from configural invariance to strict invariance. Examining different degrees of measurement invariance is commonly accomplished by employing multiplegroups factor models with increasingly severe across-group restrictions on parameters (cf. Mascherek & Zimprich, 2012; Zimprich, Allemand, & Hornung, 2006; Zimprich & Mascherek, 2010, 2011). Note that an unambiguous comparison of men and women regarding their factor means in Anger-In, Anger-Out, and Anger-Control would require at least strong measurement invariance to hold across gender (Meredith, 1993; Meredith & Horn, 2001). Hence, a first aim of the present study was to examine the degree of measurement invariance of the AX subscales of the STAXI across gender by utilizing multiple-groups confirmatory factor analysis. After having established measurement invariance, a second aim of the present study focused on differences in anger expression between female and male Swiss adolescents. Method Sample Data come from a study on bullying among students of public schools in the Canton Zug, Switzerland (cf. Willi & Hornung, 2002). The sampling procedure included a random selection of students visiting schools at the lower secondary level in the Canton Zug, Switzerland. Within schools at the lower secondary level, students of grades 7 and 9 were selected for participation, which resulted in a total sample size of N ¼ 1094 participants with a mean age of 14.6 years (SD ¼ 1.16 years, Range ¼ 13–17 years). Of these, 566 students (52%) were male and 528 were female (48%). Procedure Participants completed a questionnaire in the class room during normal school hours. Research assistants distributed the questionnaires, explained the procedure and answered students’ questions regarding the study and its aims. The questionnaire included items measuring bullying, violent delinquency, attitudes toward violence, self-esteem, sense of coherence, and questions about students’ families and peers. Part of the questionnaire were the AX subscales (Anger-In, Anger-Out, and Anger-Control) of the State-Trait Anger Expression Inventory (STAXI; Spielberger, 1988, 1996). Measures Anger expression Anger expression was measured using the Anger-In, Anger-Out, and Anger-Control subscales of the State-Trait Anger Expression Inventory (STAXI; Spielberger, 1988, 1996) in its authorized German translation (Schwenkmezger et al., 1992). The

1016

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

subscales consist of eight items each and require participants to answer self-statements (e.g., Anger-In: “I withdraw from other people”, Anger-Out: “I do things like slamming doors”, Anger-Control: “I control my anger feelings”) on a four-point Likert-type scale ranging from 1 (almost never) to 4 (almost always).1 All analyses were conducted using MPLUS version 3.0 (Muthén & Muthén, 2004), employing the maximum likelihood estimator (ML). Goodness-of-fit of the models we examined was evaluated using the Root Mean Square Error of Approximation (RMSEA), where values less than 0.06 are indicative of an acceptable model fit (cf. Browne & Cudeck, 1993), and the Standardized Root Mean Square Residual (SRMR), where values smaller than 0.07 denote an acceptable model (Hu & Bentler, 1999). As an additional criterion for absolute model fit, the c2-test is reported. In comparing the relative fit of nested models, we used the c2-difference test (Satorra & Bentler, 2001). As a measure of effect size for mean differences, we report Cohen’s d (Cohen, 1988, p. 20). Results Factorial structure of the AX scales Factor analyses started with Model M0, a single-group confirmatory three-factor model (Anger-In, Anger-Out, AngerControl), i.e., each of the 24 items was specified to load on the anger expression factor it was designated to measure. As indexed by the RMSEA (see Table 1), Model M0 achieved an acceptable model fit, whereas the SRMR was just above the critical value. The chi-square indicated statistically significant departures between the model and the data – which, however, may be due to the relatively large sample size. Factor loadings are shown in Tables 2 and 3. The average standardized factor loading was 0.57 (ranging from 0.36 to 0.73), implying that about 33% of variance were explained in the 24 items. Anger-In and AngerControl were virtually uncorrelated (r ¼ 0.01), whereas Anger-In and Anger-Out correlated moderately (r ¼ 0.18). The strongest correlation emerged for Anger-Out and Anger-Control (r ¼ 0.54), implying that those reporting to express anger externally tended to show less anger-control. Because the SRMR indicated room for model improvement, there were covariances that remained unaccounted for by Model M0. In particular, for Item 8 (“I am angrier than I am willing to admit”), which loaded on Anger-In, there were associations with both Anger-Out and Anger-Control that Model M0 did not capture adequately. Similarly, for Item 10 (“I am angrier than others are aware of”), which also loaded on Anger-In, there were associations with Anger-Control unaccounted for by Model M0. In the next Model, Model M0a, three cross-loadings (Item 8 on Anger-Out and Anger-Control, Item 10 on Anger-Control) were freely estimated. It turned out that Item 8 loaded weakly positive on Anger-Out (0.24) and weakly negative on Anger-Control (0.19). In addition, Item 10 also loaded negatively on Anger-Control (0.19). Note that similar cross-loadings have been reported in the normative sample of the German version of the AX scales (Schwenkmezger et al., 1992; cf. Kassinove, Sukhodolsky, Eckhardt, & Tsytsarev, 1997). The fit of Model M0a represented an improvement, with both the RMSEA and the SRMR falling below their critical values (see Table 1). Also, the chi-square-difference signaled that model fit had become better.2 The negative cross-loadings on Anger-Control imply that the desire to not show the intensity of one’s anger collides with the ability to control angry feelings. Similarly, the cross-loading on Anger-Out implies that those who try to not to admit intense anger are not completely successful in doing so. By contrast to Model M0, Anger-In and Anger-Control were now weakly, but significantly correlated (r ¼ 0.08), while the correlation between Anger-In and Anger-Out had attenuated slightly (r ¼ 0.13). As before, the strongest correlation emerged for Anger-Out and Anger-Control (r ¼ 0.55). To summarize, we found support for the proposed three-factor structure of the AX scales in Swiss adolescents. There were, however, three exceptions from a perfect congeneric structure, which manifested themselves in three cross-loadings. Factor correlations showed that those adolescents who considered themselves as showing more physical or verbal aggression when angry tended to also report slightly more active anger suppression and less subjective control over their angry feelings. At the same time, the small association between Anger-In and Anger-Control indicates that, in adolescence, these two ways of managing angry feelings were independent. Measurement invariance of the AX scales Next, the degree of measurement invariance of the AX scales across gender in adolescence was examined. In a first model of configural invariance (Model M1), Model M0a was estimated simultaneously in both the female and the male subsamples. As can be seen from Table 1, Model M1 achieved an acceptable fit, implying that configural invariance holds. However, in the

1 Note that the German version of the STAXI, and thus, the AX scales, consists of items that, in part, slightly differ in part from the English version (see Schwenkmezger et al., 1992). 2 Due to the comparatively low factor loadings of Items 6 and 20, we re-estimated Model M0a with these two items excluded – as suggested by an anonymous reviewer. This reduced model achieved a fit that, in terms of the stand-alone indices RMSEA and SRMR, was virtually indistinguishable from that of Model M0a (c2 ¼ 832, df ¼ 202, RMSEA ¼ 0.053, SRMR ¼ 0.057). From this we concluded that the inclusion of Items 6 and 20 did not produce a substantial misfit. Rather, Model M0a adequately accounted for the fact that both items were more weakly associated with the other items but still contributed significantly to their respective factors. We therefore decided to continue the analyses with Items 6 and 20 included. Note that Items 6 and 20 also had the lowest factor loadings in the normative samples of the German version of the AX scales (Schwenkmezger et al., 1992).

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

1017

Table 1 Fit indices for confirmatory factor models. Model

c2

df

M0 M0a M1 M1a M2 M3 M3a M4 MV MCV MMCV MMCVa

1149* 1010* 1279* 1215* 1248* 1410* 1322* 1366* 1371* 1372* 1524* 1373*

249 246 492 491 515 536 535 559 562 565 568 566

Dc2

Δdf

139*

3

64* 33* 162* 88* 44* 5* 1* 152* 1a

1 24 21 1 24 3 3 3 1a

RMSEA

90% C.I. RMSEA

SRMR

0.0572 0.0530 0.0538 0.0517 0.0508 0.0543 0.0516 0.0511 0.0510 0.0508 0.0552 0.0508

0.0538–0.0605 0.0496–0.0564 0.0502–0.0574 0.0480–0.0553 0.0472–0.0543 0.0509–0.0578 0.0481–0.0551 0.0477–0.0546 0.0476–0.0545 0.0474–0.0543 0.0519–0.0585 0.0474–0.0542

0.072 0.058 0.062 0.056 0.059 0.062 0.060 0.061 0.063 0.063 0.071 0.063

*p < .01. Note. N ¼ 1107; df ¼ degrees of freedom; Dc2 ¼ c2-difference; Δdf ¼ degrees of freedom difference; RMSEA ¼ Root Mean Square Error of Approximation; SRMR ¼ Standardized Root Mean Square Residual; M0 ¼ congeneric three-factor model (single group model); M0a ¼ modified three-factor model, including three cross-loadings (single group model); M1 ¼ Model of configural invariance; M2 ¼ model of weak invariance; M3 ¼ model of strong factorial invariance; M3a ¼ modified model of strong factorial invariance, including one variant intercept; M4 ¼ model of strict factorial invariance; MV ¼ model of equal factor variances; MCV ¼ model of equal factor covariances; MMCV ¼ model of equal factor means; MMCVa ¼ modified model of equal factor means. a Represents the difference to Model MCV.

male subsample there remained a large covariance unaccounted for between Item 24 (“I am angrier than I am willing to show”) and the factor of Anger-Control. Although the fit of Model M1 was adequate, we decided to freely estimate a crossloading of Item 24 on Anger-Control in male adolescents. This modified model of configural invariance (M1a) achieved an acceptable fit, which represented a significant improvement compared to Model M1 (see Table 1). The cross-loading of Item 24 on Anger-Control was 0.31, showing that in male adolescents the attempt to keep one’s anger in does not always succeed, because it is negatively related to anger control. We decided to accept Model M1a as the model of configural invariance. Note that this implies that complete configural invariance was absent, because there was a cross-loading of Item 24 on AngerControl in male adolescents, but not female adolescents. Thus, one may conclude that partial configural invariance, but almost complete invariance held across gender. In line with this conclusion, re-estimating Model M1a after excluding Item 24 led to a good model fit (c2 ¼ 1110, df ¼ 448, RMSEA ¼ 0.052, SRMR ¼ 0.055). Subsequently, in Model M2, factor loadings were constrained to be equal in the two subsamples while at the same time factor variances were freely estimated, thus imposing weak measurement invariance. As can be seen from Table 1, Model M2 achieved an acceptable fit, which, importantly, did not differ significantly from the fit of Model M1a. From that one may conclude that weak measurement of the AX scales held across gender. Strictly speaking, we established partial weak measurement invariance, which, however, almost represented complete weak measurement invariance. As before, we re-estimated Model M2 after exclusion of Item 24, which resulted in a good model fit (c2 ¼ 1139, df ¼ 471, RMSEA ¼ 0.051, SRMR ¼ 0.057). Next, we examined whether strong measurement invariance held across gender by constraining item intercepts to be equal, while at the same time factor means were estimated freely (Model M3). According to Table 1, doing so led to a decrement of model fit, which was both significant and substantial as judged by the RMSEA and the SRMR. It was the AngerOut Item 6 (“I exclaim threats without intending to realize them”) in particular, for which there remained an intercept difference between male and female adolescents. After freeing the intercept of Item 6 (Model M3a), an acceptable model fit was achieved (see Table 1). The intercept of Item 6 was 1.78 in males, but 1.47 in females. That is, after controlling for individual and gender differences in Anger-Out, male adolescents still had a stronger tendency to endorse Item 6. From Model M3a we concluded that partial strong invariance held. The exception of complete strong measurement invariance was Item 6. In a final step of investigating measurement invariance, we imposed strict invariance by constraining residual variances to be equal across gender (Model M4). As Table 1 shows, although model fit decreased, fit differences where marginal. From this result, we considered strict invariance to hold across gender, implying that residual variances of the 24 AX scale items were equal in both subsamples. Note that, strictly speaking, only partial strict measurement invariance was established due to the cross-loading of Item 24 in the male subsample. We re-estimated Model M4 after exclusion of Item 24, which resulted in a good model fit (c2 ¼ 1254, df ¼ 513, RMSEA ¼ 0.051, SRMR ¼ 0.059). Gender-related differences in anger expression After having established strict measurement invariance we examined gender-related differences in anger expression by constraining factor parameters to be equal. In a first model (Model MV), we constrained factor variances in the strict invariance model to be equal in both subsamples. Doing so hardly altered model fit (see Table 1), which implies that the amount of individual differences in Anger-In, Anger-Out, and Anger-Control was virtually the same. That is, within groups (male or female adolescents), individuals differed to an equal amount regarding anger expression. In the next model, in addition to equal factor variances we also constrained factor covariances to be equal (Model MCV). Note that because in Model MCV factor variances and covariances were equal, factor intercorrelations were also equal across

1018

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022 Table 2 Factor correlations on the latent level. (1) (1) Anger-Control (2) Anger-In (3) Anger-Out

0.05* 0.62*

(2)

(3)

0.08*

0.55* 0.13*

0.17*

*p < .05; Note: Correlations in the lower triangle indicate factor correlations between the latent anger factors in a model with strict measurement invariance (Model MCV); correlations in the upper right triangle indicate factor correlations in a singlegroup factor model (Model M0a).

gender. As Table 1 shows, the fit of Model MCV was practically indistinguishable from that of the previous one, implying that covariances and correlations were equal in both subsamples. Anger-In and Anger-Control were virtually unrelated (r ¼ 0.05), whereas the correlation between Anger-In and Anger-Out was weak, but statistically significant (r ¼ 0.17). As in the singlegroup model, the strongest correlation emerged for Anger-Out and Anger-Control (r ¼ 0.62). That is, in both subsamples the tendency to outwardly express anger conflicted with the tendency to control angry feelings. Finally, we also constrained factor means to be equal across gender (Model MMCV in Table 1). Doing so led to substantial decrement in model fit. We found that factor means of Anger-In were almost equal in both subsamples, whereas the other two factor means differed considerably. Consequently, in a modified model (Model MMCVa), only the factor mean of Anger-In was constrained to be equal. As Table 1 shows, this modified model achieved an acceptable fit, which, importantly, did not differ from that of Model MCV. From this we concluded that male and female adolescents only differed in Anger-Out and AngerControl. In terms of effect sizes (Cohen’s d), factor mean differences were 0.07 in Anger-In, 0.37 in Anger-Out, corresponding to a small effect which implies that female adolescents had a lower tendency of expressing anger outwardly, and 0.47 in Anger-Control, corresponding to a medium effect which implies that female adolescents had a lower anger control. Factor mean differences together with 84% confidence intervals are depicted in Fig. 1. Discussion The purpose of the present study was to investigate measurement invariance of the STAXI anger expression subscales (i.e., Anger-In, Anger-Out, and Anger-Control) and gender-related differences in a sample of Swiss adolescents. Specifically, in examining the degree of measurement invariance across sex, this article aims at establishing conditions that render the comparison of the STAXI anger expression scores meaningful across female and male adolescents. The results of the present study show that strict factorial invariance for the STAXI anger expression held across sex – with two exceptions we will turn to below. The finding of strict factorial invariance implies absence of measurement bias of the Table 3 Parameter estimates of the multiple-groups confirmatory factor analysis. Items Factor loadings 2: I tend to bottle things up. 4: I boil inside, but I don’t show it. 5: I feel angry, but I don’t tell anyone. 8: I am angrier than I am willing to admit. 10: I am angrier than others are aware of. 21: I could burst with anger, but I don’t show it. 22: I withdraw from people. 24: I am angrier than I am willing to show. 6: I exclaim threats without intending to realize them. 7: I do things like slam doors. 11: I say nasty things. 15: I get infuriated. 17: I blurt out so that others notice my anger. 18: I get upset. 19: I fly off the handle 23: I lose my poise. 1: I keep my feelings under control. 3: I keep my calm. 9: I control my behavior. 12: I can stop myself from getting angry. 13: Outwardly, I keep my countenance. 14: I try to be tolerant 16: I control my angry feelings. 20: I tell myself: Don’t get upset.

Anger-In 0.413 0.657 0.529 0.448 0.594 0.673 0.415 0.576

Note. N ¼ 1107 (531 female), Parameter estimates are based on Model MMCVm. a Only in the male subsample.

Anger-Out

Anger-Control

0.227

0.216a 0.195a

0.311a 0.367 0.501 0.432 0.535 0.649 0.636 0.712 0.634 0.695 0.709 0.672 0.522 0.597 0.538 0.645 0.347

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

1019

Fig. 1. Factor mean differences between male and female adolescents.

STAXI due to gender differences which allows for a comparison of factor means, factor variances, and factor covariances across sex that is not affected by differential functioning of the AX subscales. Using a measure that has repeatedly been shown to be subject to social desirability bias in answering behavior (Foley, Hartman, Dunn, Smith & Goldberg.. 2002; McEwan et al., 2009), it is especially important to explicitly test different degrees of measurement invariance. Although strict MI does not completely rule out the possibility that answering biases toward social desirability were present, it ensures that the questionnaire as such functions equivalently across the two groups. Note that this implies that even if social desirability had biased the answering behavior in one way or the other, it did so to the same extent in both groups. This would imply that the constructs in question (anger expression) and social desirability are indistinguishable in the sense of being methodologically inseparable within one type of analysis (i.e., CFA). A different approach would be to include either external criteria (which, however, might themselves be subject to social desirability) or a direct measure of social desirability – which then would require an analysis of MI for this measure as well. We prefer the interpretation that it is unlikely that social desirability affects both sexes to the same extent. However, even if so, this would (due to the equality of the effects) not distort sex differences in anger expression once MI is established. The two exceptions from perfect measurement invariance that occurred for Item 6 and 24 are in line with previous studies that have also found that AX scale items are not perfectly factor pure (del Barrio et al., 2004; Kassinove, Sukhodolsky, Eckhardt & Tsytsarev , 1997). However, this does not generally abrogate our conclusion of equal functioning across groups, because according to Byrne, Shavelson, and Muthén (1989), partial measurement invariance is established as long as the majority of items remains invariant (see Millsap, 2010). Also, excluding the respective items from the analyses did not alter the model fit, which can be taken as another indicator that the deviations were deniable. Notwithstanding, the two deviations from strict MI require, first, that they are studied in independent studies. Should this, second, repeatedly lead to the conclusion that they are not invariant, they should be excluded from the AX scales – at least when the two sexes are compared. Turning to the interpretation of the results on latent level, equality constraints on the variances across groups did not lead to a significant decrease in model fit. This indicates that both, female and male adolescents were comparable concerning interindividual differences in anger expression. The notion of males being generally more homogeneously grouped around the high end of aggression expression and females being homogeneously grouped around the opposite end of aggression expression (cf. Brody, 1997) is, hence, rather a result of social awareness and recognition of expected behavior than a precise description of real (self-reported) behavior variation. Empirically, the present study demonstrates that in both girls and boys the range of anger-expression is equal. Second, factor correlations were equal across groups. We found virtually no relation between Anger-In and Anger-Control, a small but significant relation between Anger-In and Anger-Out and a strong negative relation between Anger-Out and Anger-Control, which closely corresponds to the finding from the normative sample of the German version of the AX scales (Schwenkmezger et al., 1992). The small relation between Anger-In and Anger-Out indicates that both types of anger expression are not opposite poles of a continuum but are different ways of dealing with anger. This result is in line with previous studies (Spielberger et al., 1985), which also found no evidence for the original idea of Anger-In and -Out as being opposites. Anger-In and -Out can be interpreted as interindividual differences in involuntary reactions to a strong emotional arousal. This interpretation is in line with findings by Lutwak et al. (2001), who found shame-proneness and guilt-proneness being related to less “visible” anger expression (Anger-In). Hence, Anger-In and -Out might demark the dispositional, more impulsive, non-cognitively controlled way of dealing with anger. Depending on temperament, individuals tend to express anger more inwardly or outwardly (cf. Eisenberg, Fabes, Nyman, Bernzweig, & Pinuelas, 1994; Ortiz & del Barrio, 2006). The

1020

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

small relation that was found between the two subscales clearly indicates that inferences from one type of anger expression to the total level of anger expression are not warranted. Turning to the relation between Anger-Control and the two other subscales, one might argue that Anger-Control describes a more cognitive way to manage anger that is influenced by experience. Control could, hence, be looked at as a superordinate ability in managing anger. Anger-Control can, independently from the tendency to express anger inwardly or outwardly, be applied as a rather cognitive, self-regulatory strategy to cope efficiently and appropriately with anger. The strong, negative relation between Anger-Out and Control shows that adolescents who heavily rely on Anger-Out do not rely much on AngerControl or, alternatively, are unable to use it effectively (cf. Kassinove et al., 1997). Hence, we conclude that Anger-Out does not seem to be part of the repertoire an individual applies when having expertise in anger management. It rather could be a way of anger expression that is an obstacle in learning to deal or dealing with anger in a controlled way (see Brody, 1985). However, Anger-In is independent from Control and, hence, does neither foster nor prevent successful cognitive control strategies. Anger-In might represent a “hot” impulsive reaction to anger that is socially acceptable because it does not cause disturbances in the (social) environment of the angry individual. Adolescents of the age of participants in the present sample have already learned conventions and acceptable expressions of emotions (e.g., Parker et al., 2001). Emotion expression has been shown to be influenced by social context (e.g., Evers et al., 2005). Suggesting that Anger-Out is less accepted by society implies a need of anger control strategies. By contrast, Anger-In does not appear to be in the need of control as much because it does not obviously harm others and, thus, appears less malevolent. Similarly, Schwenkmezger et al. (1992) argued that attempts to control anger represent a socially accepted and expected way of dealing with anger and, hence, typically aim at suppressing the outward expression of anger – as opposed to Anger-In, which can go unnoticed by others. An alternative interpretation would consider successful Anger-Control as mainly involving anger suppression and hence consider AngerControl and Anger-Out as opposites. Kopper and Epperson (1996), for example, have shown that in a comprehensive factor analysis of instruments tapping anger or anger expression, Anger-Control and Anger-Out loaded on the same factor – but with opposite signs. Anger-In, by contrast, shared large amounts of variance with suspicion and irritability, but not with Anger-Control and Anger-Out. Turning to the interpretation of mean levels, we found lower levels of Anger-Control in girls than in boys – even after having established strict MI. Girls also showed lower levels of Anger-Out (although the effect was small) whereas they were about the same for Anger-In. These results are in line with sex differences reported in the German manual of the STAXI, although in the normative sample the effect size of the sex difference in Anger-Control was smaller (Schwenkmezger et al., 1992). One could argue that the lower levels of Anger-Control and Anger-Out in girls indicate that because less Anger-Out is present, the necessity to control anger is less given. When women are angry they appear to be more sensitive to the feelings of others, which might account for a suppression of an outward expression of anger by controlling anger (Evers et al., 2005). What appears puzzling, however, is that one would expect that exactly for this reason women would need more, not less, control of anger. Alternatively, as Kring and Gordon (1998), women’s greater experience in general with emotions may help them to also deal with anger more effectively, implying that they need less control than men when experiencing the same level of anger. Note that when situational information is provided in self-report measures, or when participants are interviewed about angry feelings, women actually express more anger in terms of frequency and intensity than men do. Moreover, women also report more enduring experiences of anger than do men (cf. Brody & Hall, 2000). This seemingly contradiction underscores that anger expression and the experience of anger are not necessarily perfectly related. In a review on gender and emotion, Brody (1997) extensively discussed research that showed gender differences in the expression modality of anger – but not for anger experience. Hence, we conclude that the lower levels of Anger-Out in girls could mirror socialization, because outwardly expressing anger seems to be less acceptable for girls, and should not to be interpreted as lower levels of anger experience (see also Evers et al., 2005). Because inwardly expressing anger is less prone to social appraisal, girls and boys did not differ on this dimension. To summarize, what do the results of the present study reveal about anger expression in Swiss adolescents? First, concerning the measurement properties of the STAXI we found (with two exceptions) strict measurement invariance, which indicates that the questionnaire functions equally in boys and girls. Second, we found that variances of the anger expression subscales were equal across groups, indicating that the range of interindividual differences in anger expression is equivalent. Third, the factor correlations were equal across groups. And suggest that Anger-Control seems to indicate a cognitive aspect of managing anger expression, whereas Anger-In and Anger-Out demark the rather intuitive aspects of anger expression. Fourth, girls report lower levels of Anger-Out and Anger-Control but no differences in Anger-In occurred between girls and boys on average. Acknowledgment The authors would like to thank Rainer Hornung from the Institute of Psychology at the University of Zurich for sharing the data with us. References Armstead, C. A., & Clark, R. (2002). Assessment of self-reported anger expression in pre- and early-adolescent African Americans: psychometric considerations. Journal of Adolescence, 25, 365–371. Averill, J. R. (1982). Anger and aggression: An essay on emotion. New York: Springer. Averill, J. R. (1983). Studies on anger and aggression: implications for theories of emotion. American Psychologist, 38, 1145–1160.

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

1021

del Barrio, V., Aluja, A., & Spielberger, C. D. (2004). Anger assessment with the STAXI-CA: psychometric properties of a new instrument for children and adolescents. Personality and Individual Differences, 37, 227–244. Berkowitz, L. (1999). Anger. In T. Dalgleish, & M. Power (Eds.), Handbook of cognition and emotion (pp. 411–428). Chichester: Wiley. Bongard, S., & al’Absi, M. (2003). Domain-specific anger expression assessment and blood pressure during rest and acute stress. Personality and Individual Differences, 34, 1383–1402. Brody, L. R. (1985). Sex differences in emotional development: a review of theories and research. Journal of Personality, 53, 102–149. Brody, L. R. (1997). Gender and emotion: beyond stereotypes. Journal of Social Issues, 53, 369–394. Brody, L. R., & Hall, J. A. (2000). Sex, emotion, and expression. In M. Lewis, & J. M. Haviland-Jones (Eds.), Handbook of emotions (pp. 338–349). New York: Guilford. Browne, M. W., & Cudeck, R. (1993). Alternative ways of model fit. In K. A. Bollen &J, & S. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park: Sage. Byrne, B. M., Shavelson, R. J., & Muthén, B. (1989). Testing for the equivalence of factor covariance and mean structures: the issue of partial measurement invariance. Psychological Bulletin, 105, 456–466. Campbell, A., & Muncer, S. (1987). Models of anger and aggression in the social talk of women and men. Journal for the Theory of Social Behaviour, 17, 489–511. Cautin, R. L., Overholser, J. C., & Goetz, P. (2001). Assessment of mode of anger expression in adolescent psychiatric inpatients. Adolescence, 36, 163–170. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Crandall, V. C., Crandall, V. J., & Katkovsky, W. (1965). A children’s social desirability questionnaire. Journal of Consulting Psychology, 29, 27–36. Davies, M. F. (2001). Socially desirable responding and impression management in the endorsement of love styles. Journal of Psychology, 135, 562–570. Deffenbacher, J. L., Oetting, E. R., Lynch, R. S., & Morris, C. D. (1996). The expression of anger and its consequences. Behaviour Research and Therapy, 34, 575–590. Eagly, A. H., & Steffen, V. J. (1986). Gender and aggressive behavior: a meta-analytic review of the psychosocial literature. Psychological Bulletin, 100, 309–330. Eckhardt, C., Jamison, T. R., & Watt, K. (2002). Anger experience and expression among male dating violence perpetrators during anger arousal. Journal of Interpersonal Violence, 17, 1102–1114. Eftekhari, A., Turner, A. P., & Larimer, M. E. (2004). Anger expression, coping, and substance use in adolescent offenders. Addictive Behaviors, 29, 1001–1008. Eisenberg, N., Fabes, R. A., Nyman, M., Bernzweig, J., & Pinuelas, A. (1994). The relations of emotionality and regulation to children’s anger-related reactions. Child Development, 65, 109–128. Ekman, P. (1992). Facial expression and emotion. American Psychologist, 48, 384–392. Evers, A., Fischer, A. H., Rodriguez Mosquera, P. M., & Manstead, A. S. R. (2005). Anger and social appraisal: a “spicy” sex difference? Emotion, 5, 258–266. Fassino, S., Abbate-Daga, G., Pierò, A., Leombruni, P., & Rovera, G. G. (2003). Dropout from brief psychotherapy within a combination treatment in bulimia nervosa: role of personality and anger. Psychotherapy and Psychosomatics, 72, 203–210. Fischer, A. H., & Manstead, A. S. R. (2000). The relation between sex and emotions in different cultures. In A. H. Fischer (Ed.), Sex and emotion: Social psychological perspectives (pp. 71–94). Cambridge: Cambridge University Press. Fisher, P. C., Smith, R. J., Leonard, E., Fuqua, D. R., Campbell, J. L., & Masters, M. A. (1993). Sex differences on affective dimensions: continuing examination. Journal of Counseling and Development, 71, 440–443. Foley, P. F., Hartman, B. W., Dunn, A. B., Smith, J. E., & Goldberg, D. M. (2002). The utility of the state-trait anger expression inventory with offenders. International Journal of Offender Therapy and Comparative Criminology, 46, 364–378. Forgays, D. G., Forgays, D. K., & Spielberger, C. D. (1997). Factor structure of the state-trait anger expression inventory. Journal of Personality Assessment, 69, 497–507. Funkenstein, D. H., King, S. H., & Drolette, M. (1954). The direction of anger during a laboratory stress-inducing situation. Psychosomatic Medicine, 16, 404–413. Fuqua, D. R., Lenard, E., Masters, M. A., Smith, R. J., Campbell, J. L., & Fischer, P. C. (1991). A structural analysis of the state-trait anger expression inventory. Educational and Psychological Measurement, 51, 439–446. Furlong, M. J., & Smith, D. C. (1994). Anger, hostility, and aggression: Assessment, prevention, and intervention strategies for youth. New York: Wiley. Ganster, D., Hennessey, H., & Luthans, F. (1983). Social desirability response effects: three alternative models. Academy of Management Journal, 26, 321–331. Gerhards, F. (1992). Inhibition of anger expression in migraine patients. Zeitrschrift Für Klinische Psychologie, 21, 262–271. Hess, U., & Kirouac, G. (2000). Emotion expression in groups. In M. Lewis, & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed.). (pp. 368–381) New York: The Guilford Press. Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement invariance in aging research. Experimental Aging Research, 18, 117–144. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. Kassinove, H., Sukhodolsky, D. G., Eckhardt, C. I., & Tsytsarev, A. V. (1997). Development of a Russian state-trait anger expression inventory. Journal of Clinical Psychology, 53, 543–557. Kopper, B. A., & Epperson, D. L. (1996). The experience and expression of anger: relationships with gender, gender role socialization, depression, and mental health functionning. Journal of Counseling Psychology, 43, 158–165. Kring, A. M. (2000). Sex and anger. In A. H. Fischer (Ed.), Sex and emotions: Social psychological perspectives (pp. 211–231). Cambridge: Cambridge University Press. Kring, A. M., & Gordon, A. H. (1998). Sex differences in emotion: expression, experience, and physiology. Journal of Personality and Social Psychology, 74, 686–703. Lutwak, N., Panish, J. B., Ferrari, J. R., & Razzino, B. E. (2001). Shame and guilt and their relationship to positive expectations and anger expressiveness. Adolescence, 36, 641–653. Mascherek, A., & Zimprich, D. (2012). Age-related differences in typical intellectual engagement in young and old adults. Experimental Aging Research, 38, 63–86. McEwan, T. E., Davis, M. R., MacKenzie, R., & Mullen, P. E. (2009). The effects of social desirability response bias on STAXI-2 profiles in a clinical forensic sample. British Journal of Clinical Psychology, 48, 431–436. McKenna, M. C., Zevon, M. A., Corn, B., & Rounds, J. (1999). Psychological factors and the development of breast cancer: a meta-analysis. Health Psychology, 18, 520–531. Meredith, W. (1993). Measurement invariance, factor analysis, and factorial invariance. Psychometrika, 58, 525–543. Meredith, W., & Horn, J. L. (2001). The role of factorial invariance in modeling growth and change. In L. M. Collins, & A. G. Sayer (Eds.), New methods for the analysis of change (pp. 203–240). Washington DC: American Psychological Association. Millsap, R. (2010). Testing measurement invariance using item response theory in longitudinal data: an introduction. Child Development Perspectives, 4, 5–9. Musante, L., & Treiber, F. A. (2000). The relationship between anger-coping styles and lifestyle behaviors in teenagers. Journal of Adolescent Health, 27, 63–68. Musante, L., Treiber, F. A., Davis, D. C., Waller, J. L., & Thompson, W. O. (1999). Assessment of self-reported anger in youth. Assessment, 6, 225–233. Muthén, L. K., & Muthén, B. (2004). MPLUS user’s guide. Los Angeles, CA: Muthén & Muthén. Ortiz, M. ÁC., & del Barrio Gándara, V. (2006). Study on the relations between temperament, aggression, and anger in children. Aggressive Behavior, 32, 207–215. Parker, E. H., Hubbard, J. A., Ramsden, S. R., Relyea, N., Dearing, K. F., Smithmyer, C. M., et al. (2001). Children’s use and knowledge to display rules for anger following hypothetical vignettes versus following live peer interaction. Social Development, 10, 528–557. Plant, A. E., Hyde, J. S., Keltner, D., & Devine, P. G. (2000). The sex stereotyping of emotions. Psychology of Women Quarterly, 24, 81–92. Plant, E. A., Kling, K. C., & Smith, G. L. (2004). The influence of sex and social role on the interpretation of facial expressions. Sex Roles, 51, 187–197. Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: a critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879–903. Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66, 507–514.

1022

D. Zimprich, A. Mascherek / Journal of Adolescence 35 (2012) 1013–1022

Schimmack, U. (2002). Frequency judgements of emotions: the cognitive basis of personality assessment. In P. Sedlmeier, & T. Betsch (Eds.), Frequency processing and cognition (pp. 189–204). New York, NY, US: Oxford University Press. Schwenkmezger, P., Hodapp, V., & Spielberger, C. D. (1992). Das state-Trait-Ärgerausdrucks-Inventar (STAXI). Bern: Huber. Sharkin, B. S. (1993). Anger and sex: theory, research, and implications. Journal of Counseling and Development, 71, 386–389. Spielberger, C. D. (1988). Manual for the state trait anger expression inventory (STAXI). Odessa, FL: Psychological Assessment Resources. Spielberger, C. D. (1996). State-trait anger expression inventory: Professional manual. Odessa, FL: Psychological Assessment Resources. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. J., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: construction and validation of an anger expression scale. In M. A. Chesney, & R. H. Rosenman (Eds.), Anger and hostility in cardiovascular and behavioural disorders (pp. 5– 30). Washington, DC: Hemisphere Publishing Corporation. Stoner, S. B., & Spencer, W. B. (1987). Age and sex differences with the anger expression scale. Educational and Psychological Measurement, 47, 487–492. Suter, J. M., Byrne, M. K., Byrne, S., Howells, K., & Day, A. (2002). Anger in prisoners: women are different from men. Personality and Individual Differences, 32, 1087–1100. Timmers, M., Fischer, A. H., & Manstead, A. S. R. (1998). Gender differences in motives for regulating emotions. Personality and Social Psychology Bulletin, 24, 974–985. Willi, M., & Hornung, R. (2002). Jugend und Gewalt – Ergebnisse einer Befragung von Schülerinnen und Schülern im Kanton Zug. Bern: Lang. Zimprich, D., Allemand, M., & Hornung, R. (2006). Measurement invariance of the abridged sense of coherence scale in adolescents. European Journal of Psychological Assessment, 22, 280–287. Zimprich, D., & Mascherek, A. (2010). Five views of a secret: does cognition change during middle adulthood? European Journal of Ageing, 7, 135–146. Zimprich, D., & Mascherek, A. (2011). Measurement invariance and age-related differences of trait anger across the adult lifespan. Personality and Individual Differences, . doi:10.1016/j.paid.2011.10.030.