Journal of Homosexuality

ISSN: 0091-8369 (Print) 1540-3602 (Online) Journal homepage: http://www.tandfonline.com/loi/wjhm20

Assessing the Gender Invariance of the Modern Homonegativity Scale Daniel H. Romero MA, Osvaldo F. Morera PhD & John S. Wiebe PhD To cite this article: Daniel H. Romero MA, Osvaldo F. Morera PhD & John S. Wiebe PhD (2015) Assessing the Gender Invariance of the Modern Homonegativity Scale, Journal of Homosexuality, 62:11, 1539-1559, DOI: 10.1080/00918369.2015.1073034 To link to this article: http://dx.doi.org/10.1080/00918369.2015.1073034

Accepted online: 16 Jul 2015.Published online: 16 Jul 2015.

Submit your article to this journal

Article views: 32

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=wjhm20 Download by: [University of Lethbridge]

Date: 28 September 2015, At: 21:05

Journal of Homosexuality, 62:1539–1559, 2015 Copyright © Taylor & Francis Group, LLC ISSN: 0091-8369 print/1540-3602 online DOI: 10.1080/00918369.2015.1073034

Assessing the Gender Invariance of the Modern Homonegativity Scale DANIEL H. ROMERO, MA

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Department of Psychology, University of North Texas, Denton, Texas, USA

OSVALDO F. MORERA, PhD and JOHN S. WIEBE, PhD Department of Psychology, University of Texas at El Paso, El Paso, Texas, USA

The measurement invariance of the Modern Homonegativity Scale (MHS) was examined among heterosexual female (n = 449) and male (n = 329) university students who were predominantly Mexican American. The MHS demonstrated full invariance of factor loadings and partial invariance of latent intercepts. At the latent mean level, heterosexual men compared to heterosexual women held more negative attitudes toward both gay men and lesbian women. There were no latent mean differences in attitudes toward gay men and lesbian women when rated by either heterosexual men or heterosexual women. The MHS can be used by heterosexual men and heterosexual women to assess their homonegativity. KEYWORDS gay men, gender invariance, lesbian women, modern homonegativity

Homonegativity refers to negative affect, perceptions, and actions directed toward individuals who are discerned, accurately or inaccurately, as being gay or lesbian (Morrison & Morrison, 2011). Modern homonegativity has more to do with ideas of equality and social impartiality than with morality (Morrison & Morrison, 2003; Satcher & Legget, 2006, 2007). The 12-item Modern Homonegativity Scale (MHS) was developed as a measure of modern prejudice toward gay males and lesbian women (Morrison & Morrison, 2003). The scale includes items such as “Lesbians (gay men) do not have all the rights they need” or “Gay men (lesbians) still need to

Address correspondence to Osvaldo F. Morera, Department of Psychology, University of Texas at El Paso, 500 W. University Ave., El Paso, TX 79968, USA. E-mail: [email protected] 1539

1540

D. H. Romero et al.

protest for equal rights,” characterized as modern measures of sexual prejudice since these items focus on changing the “status quo” (Morrison & Morrison, 2003, p. 18). The MHS has two forms, with similarly worded items involving either a male target (the MHS-G) or a female target (MHS-L). Scores from the MHS were recently found to correlate with those from an older measure of homonegativity, the Attitudes Toward Lesbians and Gay Men scale (ATLG; Herek, 1984a), demonstrating convergent validity (Rye & Meaney, 2010).

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Purpose of This Study Many studies have examined and proposed reasons why there are sex differences in attitudes toward gay men and lesbian women. While we will review some of those findings, no study has taken into account whether homonegativity scales themselves measure the same construct equally for both heterosexual males and heterosexual females. To make group comparisons on any measure, it is necessary to establish the measurement invariance of the measure. In other words, one must determine whether a set of indicators measure the same construct among different groups (Kline, 2011). It is well established in the psychometric literature that the presence or absence of group differences at the level of the observed scores does not necessarily translate into differences at the latent variable, or “true” level (Thissen, Steinberg, & Gerrard, 1986). Thus, meaningful comparisons of group differences must take place at the latent variable level (Schmitt & Kuljanin, 2008). As discussed later, multiple forms of invariance need to be established to ascertain “true” group differences at the latent variable level. Reasons for sex differences with respect to homonegativity are first discussed. Then we describe some conflicting studies that have examined observed score differences of homonegativity as a function of participant sex.

Gender Roles of Masculinity and Femininity Related to Male Homonegativity While psychometric and methodological issues related to the assessment of homonegativity are important and understudied, it is clear that gender belief systems are responsible for sex differences in attitudes toward homosexuality (Kite & Whitley, 1996). Fagot (1985) also noted that gender roles for boys are enforced more by parents than are gender roles that pertain to girls. Due to these demands, heteronormative men may avoid expressions of femininity (Kite & Deaux, 1987; McCreary, 1994). This avoidance of femininity may also be related to devaluing attitudes toward females. Some heteronormative men have negative attitudes toward gay men and heterosexual women (Kilianski,

Invariance of the Modern Homonegativity Scale

1541

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

2003), given that femininity is stereotypically related to gay men (Herek, 1984b) and women. Indeed, many heterosexual men are concerned about being perceived as feminine (Theodore & Basow, 2000) or gay (Barron, Struckman-Johnson, Quevillon, & Banka, 2008), and that concern is related to homonegativity (Wilkinson, 2004). Finally, Lehavot and Lambert (2007) found that individuals with increased homonegativity have negative feelings toward “double violators” (p. 289), or individuals who are perceived to violate expectations regarding both sexuality and gender roles. Given this theoretical justification for increased heterosexual male homonegativity (relative to that of heterosexual females), numerous researchers have examined these sex differences empirically.

Discrepancies in Past Work on Homonegativity as a Function of Rater Sex Many studies have demonstrated sex differences related to homonegativity toward gay men and lesbian women at the observed level without taking into account the differential functioning of the measure (Meade & Wright, 2012). For example, using the ATLG and other earlier conceptualized measures of homonegativity among college students, heterosexual males’ attitudes tend to be more negative toward gay men than toward lesbian women, whereas heterosexual women’s attitudes toward gay males and lesbian women are more similar or consistent (Herek, 1988; LaMar & Kite, 1998; Louderback & Whitley, 1997; VanderStoep & Green, 1988; Whitley, 1988). Conversely, when a more modern measure of homonegativity is used, heterosexual men report similar homonegativity toward gay men and lesbian women (Morrison & Morrison, 2003). Recently, Morrison, Morrison, and Franklin (2009) investigated the joint factor structure of the MHS and the ATLG among Canadian and U.S. college students. A one-factor model, which assumed the items from both the ATLG and the MHS loaded on the same factor, was tested against a two-factor solution, where ATLG items loaded on one factor and MHS items loaded on a second factor. The two-factor solution showed the best fit with the data (Morrison et al., 2009). Thus, there are differences between modern homonegativity and earlier conceptualizations of homonegativity and these differences may contribute to the inconsistent findings of previous research. While most of the above-mentioned studies were conducted in the United States, studies using Chilean and Canadian undergraduate students found that heterosexual females had more negative attitudes toward gay men than toward lesbian women using the ATLG (Cárdenas & Eduardo Barrientos, 2008; Mohipp & Morry, 2004). In a sample of Canadian undergraduates that

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

1542

D. H. Romero et al.

used the MHS, Morrison and Morrison (2003) found that heterosexual females had similar levels of prejudice toward gay men and lesbian women. Among noninstitutionalized U.S. civilian populations over the age of 14, the literature shows that heterosexual males, compared to heterosexual females, report greater negative attitudes toward gay men (Aberson, Swan, & Emerson, 1999; Cotton-Huston & Waite, 2000; Herek, 1988; Louderback & Whitley, 1997; Marsiglio, 1993; Morrison & Morrison, 2003, 2011; Morrison et al., 2009), regardless of the measure used to assess homonegativity. In addition, when using the MHS, heterosexual men compared to heterosexual women also reported greater levels of homonegativity toward lesbian women among samples of Canadian undergraduates (Morrison & Morrison, 2003), nonstudent Canadian university employees (Morrison & Morrison, 2011), and American undergraduates (Morrison et al., 2009). Conversely, using earlier measures of homonegativity, studies demonstrated that heterosexual men compared to heterosexual women rated lesbian women less negatively (Herek, 1988; Louderback & Whitley, 1997; Whitley, 1988). Thus, while the ATLG appears sensitive to differences between cultures, the MHS demonstrates stable sex differences across cultures. In summary, discrepancies in homonegativity sex differences may be due to a variety of factors, including the measure used, cultural differences across samples, and the type of sample. While there are conflicting findings, no study has examined whether these sex differences occur at the latent variable level. Such an examination is critical, as group differences at the observed score level may not be present at the latent variable level. While an earlier study examined differences in factor loadings across groups (Morrison et al., 2009), no study has examined sex differences on any measure of homonegativity at the latent variable level.

Hypotheses In this study, we sought to examine the gender invariance of the MHS. We hypothesized that the MHS scores would be measurement invariant across rater sex and across the sex of rated targets (i.e., gay men and lesbian women). Given the previous research involving the MHS (Morrison & Morrison, 2003; Morrison et al., 2009) and the reviewed literature on gender belief systems, we further hypothesized differences at the latent mean level, such that (1) heterosexual men compared to heterosexual women would express more prejudice toward gay men and lesbian women (Morrison & Morrison, 2003), and (2) heterosexual men would express more prejudice toward gay men than toward lesbian women (Herek, 1988; LaMar & Kite, 1998), whereas heterosexual women would

Invariance of the Modern Homonegativity Scale

1543

express the same amount of prejudice toward gay men and lesbian women (Herek, 1988; Morrison & Morrison, 2003).

METHOD

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Participants Participants were 853 introductory psychology students from a university in Texas who participated in the study for course credit. Of the participants, 789 self-identified as heterosexual, 36 self-identified as gay or lesbian, 26 self-identified as bisexual, and two did not report their orientation. Bisexual, lesbian, and gay participants and those who did not report their orientation were not included in the analysis since the focus was heterosexuals’ attitudes toward gay men and lesbian women. The ages of the heterosexual participants ranged from 17 to 47 years, with a mean of 20 (SD = 3.64). Participants were randomly assigned to complete the MHS assessing gay men as the target group or lesbian women as the target group. Of the participants, 79.7% were Mexican American, 6.4% were of Mexican nationality, 6.8% were Anglo American, 2.4% were African American, 1.8% were Asian American, and 2.8% self-identified as “Other.”

Measures DEMOGRAPHIC ITEMS A demographic questionnaire assessed sexual orientation, sex, belief that homosexuality is a choice or beyond an individual’s control, religious activity, whether the participant had a gay/lesbian friend, ethnicity, income, age, and political affiliation. Sexual orientation was measured using a one-item categorical variable including three responses (e.g., homosexual, heterosexual, bisexual).

MODERN HOMONEGATIVITY SCALE (MHS; MORRISON & MORRISON, 2003) Participants were asked to rate each of 12 items on a scale of 1 (strongly disagree) to 5 (strongly agree). Total scores range from 12 to 60, with higher scores indicating greater homonegativity. The reliability of the MHS-G is good, with a coefficient alpha of .91 for both male and female raters in past research (Morrison & Morrison, 2003) and .88 (95% CI: .85, .90) for both in the current study. The coefficient alpha of the MHS-L was .89 for male raters and .85 for female raters in the Morrison and Morrison (2003) study and .85 (95% CI: .82, .88) for male raters and .88 (95% CI: .86, .90) for female raters in the current study.

1544

D. H. Romero et al.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Procedures The study was approved by the Institutional Review Board for Human Subjects Research, and all participants rendered informed consent. All participants were able to decline the research invitation and were also allowed to discontinue the study at any time without penalty. No participant refused to take part in this study. Each participant had an equal chance of receiving either the MHS-L or MHS-G. Scales and questionnaires were administered in the following order: demographic questionnaire, the Family Attitude Scale (Ramirez & Carrasco, 1996), ATLG (Herek, 1984a), Marlowe Crowne Social Desirability Scale (Crowne & Marlowe, 1960), and the MHS-L or MHS-G. It is important to note that administering the ATLG before the MHS may impact MHS scores (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). After administration of the questionnaires, participants were debriefed and informed that the investigators were assessing modern prejudice toward gay men and lesbian women. The time to complete the informed consent, questionnaires, and debriefing was approximately 30–45 minutes.

Missing Data NORM (Schafer, 1999) software was used to impute data using an expectation-maximization (EM) algorithm, which follows an iterative process until estimates converge (Enders, 2010). Of the 789 remaining heterosexual participants, four men did not complete any items of the MHS and were dropped from the analysis. In addition, six women did not complete any of the MHS items and were also dropped. One female participant completed only six MHS items, but the NORM solution would not converge unless this participant was omitted. Of the participants who rated gay men, 2% of men and 1% of women skipped at least one item. Of the participants who rated lesbian women, 2% of men and 2% women skipped at least one item. Final estimates were rounded to the nearest observed value.

Assessment of Model Fit As this project involved the assessment of different forms of measurement invariance, the investigation of model fit was a critical component. To address non-normality in the data, an asymptotic variance–covariance matrix for each group was estimated in PRELIS 2.8 (Jöreskog & Sörbom, 2003). The resulting matrix was analyzed in LISREL 8.80 (Jöreskog & Sörbom, 2006), and the Satorra-Bentler scaled chi-square statistic (Satorra & Bentler, 1994) was reported. As models will almost never perfectly describe the data (MacCallum, 2003), fit indices were used to evaluate the models. An absolute fit index was

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Invariance of the Modern Homonegativity Scale

1545

assessed using standardized root mean square residual (SRMR). A parsimonious index was measured by using the root mean square error of approximation (RMSEA) and comparative fit index (CFI). Proposed guidelines for adequate fit for these indices are SRMR ≤ .08 and RMSEA < .06 (Hu & Bentler, 1999). McDonald (1999) has also proposed RMSEA ≤ .08. In addition, Blackburn, Donnelly, Logan, and Renwick (2004) suggest that a CFI greater than .90 indicates adequate fit. Finally, we also report the Non-Normed Fit Index (NNFI), where values in excess of .90 are minimally acceptable (Brown, 2006). While the SRMR and RMSEA were reported, it is also true that various forms of invariance models are nested models. To compare the parsimony of these nested models, differences between the CFI index (Cheung & Rensvold, 2001) and the differences in RMSEA were examined (Chen, 2007). Based on simulation studies, differences between nested models in magnitude greater than or equal to 0.01 on the CFI or differences in magnitude between nested models greater than or equal to 0.015 on the RMSEA statistic would be indicative of differences in model parsimony. Therefore, the CFI statistic is also reported.

Measurement Invariance Across Sex of the Participant The first step in documenting measurement invariance is to establish configural invariance, which requires the same pattern of factor loadings across all groups (Horn, McArdle, & Mason, 1983). This is the first model, within which stricter and constrained models are later assessed. Once configural invariance is established, weak and strong invariance are assessed. Weak invariance is established by constraining the factor loadings to equality across groups. This parsimony of the weak invariance model is compared to the configural invariance model using the differences in the CFI and RMSEA statistics between these models. If the weak invariance model is more parsimonious than the configural invariance model, our next step is to determine whether the strong invariance model is more parsimonious than the weak invariance model. Recall that strong invariance requires both factor loadings and latent item intercepts be constrained to equality across groups (Horn et al., 1983), while the weak invariance model constrains only factor loadings to equality across groups. The strong invariance model and the weak invariance model are compared using the differences in the CFI and RMSEA statistics. If some of the latent intercepts differ across groups but the majority of latent intercepts do not differ, we can say that partial measurement invariance of the intercepts holds (Steenkamp & Baumgartner, 1998). If partial or strong invariance holds (Steenkamp & Baumgartner, 1998), one can then determine whether males and females differ in homonegativity at the latent mean level.

1546

D. H. Romero et al.

While other forms of invariance can be established (e.g., invariance of item unique variances and invariance of the factor variances), our primary interest involves assessing weak and strong measurement invariance, as they are necessary conditions that will allow for the examination of latent mean differences across groups. If the strong invariance model is the most parsimonious model, latent mean differences are examined across the two comparison groups. In addition, Cohen’s d is reported to quantify the magnitude of the group differences at the latent variable level.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Power Analysis For the power analysis, α was set at .05, and β was set at .20. The analysis was based on a test of “not close fit” (MacCallum, Browne, & Sugawara, 1996, p. 136). In a test of “ close fit” (p. 134), it is customary to set Ro = .05, where Ro is the hypothesized value of the population value for the root mean square error of approximation (RMSEA) under the null hypothesis (MacCallum et al., 1996). The alternative hypothesis specifies that the population value of the RMSEA statistic equals .08. Using Preacher and Coffman’s (2006) software, it was determined that a minimum sample of 294 participants across both groups would be needed to test the configural invariance model. As required sample size decreases with increases in model complexity (MacCallum et al., 1996), other forms of measurement invariance require fewer participants. Since the smallest of the two groups in the present study combine to 333 individuals, there was adequate sample size to test models of configural invariance. The single common factor model was estimated for all groups prior to testing the various forms of variance.

RESULTS Descriptive Statistics Participant characteristics are provided in Table 1. MHS scores were not significantly correlated with social desirability scores (rs ranging from −.12 to .10) or age (rs ranging from −.12 to .08). MHS scores were positively correlated with total scores on the Family Attitude Scale (rs ranging from 0.35 to 0.40) A 2 (participant sex) × 2 (target sex) analysis of variance was conducted on the MHS scores. There was a main effect for participant sex, F(1, 774) = 34.89, p < .001, partial η2 = 0.043. Across levels of target sex, men displayed more homonegativity (M = 37.04, SD = 9.02) than did women (M = 33.22, SD = 8.90). There was no main effect for target sex, F(1, 774) = 0.48, p = .49, partial η2 = 0.001, nor was there an interaction between target sex and participant sex, F(1, 774) = 2.41, p = .12, partial η2 = 0.003. As

1547

Invariance of the Modern Homonegativity Scale TABLE 1 Characteristics of participants (N = 789)

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Characteristic Sex Female Male Age 17–20 21–30 31–40 41–50 Political affiliation Liberal Conservative Moderate Sexual orientation etiology belief Choice Beyond an individual’s control Contact Has a gay male friend Does not have a gay male friend Has a lesbian friend Does not have a lesbian friend Ethnicity/Nationality Mexican American Mexican Anglo American African American Asian American Other Political party Republican Democrat Independent None Cannot vote in the U.S. Religious activity Very active Moderately active Somewhat active Minimally active Not active

n (%) 449 (57.7) 329 (42.3) 595 163 13 6

(76.6) (21) (1.7) (.7)

264 (34.2) 161 (20.9) 346 (44.9) 443 (61.1) 282 (38.9) 646 132 560 218

(83.0) (17.0) (72.0) (28.0)

619 50 53 19 14 22

(79.7) (6.4) (6.8) (2.4) (1.8) (2.8)

146 332 31 176 87

(18.9) (43.0) (4.0) (22.8) (11.3)

90 182 207 159 139

(11.6) (23.4) (26.6) (20.5) (17.9)

the analyses involving measurement invariance were performed in a pairwise fashion, we also report a series of independent measures t-tests, accompanied by Cohen’s d, to quantify group differences on a pairwise basis. Regarding attitudes toward gay men and lesbian women, heterosexual males’ (M = 37.77, SD = 9.25) level of modern homonegativity was significantly greater than heterosexual females’ (M = 32.93, SD = 9.01) level of modern homonegativity on the MHS-G, t(379) = 5.15, p < .01, Cohen’s d = .53. Heterosexual males’ (M = 36.31, SD = 8.76) level of modern homonegativity was also significantly greater than heterosexual females’ (M = 33.48,

1548

D. H. Romero et al.

SD = 8.81) level of modern homonegativity on the MHS-L, t(395) = 3.16, p < .01, Cohen’s d = .32. There were no significant differences in heterosexual males’ modern homonegativity toward gay men and lesbian women and no significant differences in heterosexual females’ modern homonegativity toward gay men and lesbian women.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Modifications Established for Configural Invariance Across All Groups After examining all the residuals and local fit information for each group separately, model fit was improved when the correlated errors of items 7 and 8 were estimated across all group comparisons in a configurally invariant model. Item 7 states, “Gay men (lesbians) should stop shoving their lifestyles down other people’s throats,” and item 8 states, “If gay men (lesbians) want to be treated like everyone else, then they need to stop making such a fuss about their sexuality/culture,” and error terms of these items were allowed to correlate since the items are not particularly subtle. However, a limitation of freeing correlated errors is that model fit may be enhanced due to capitalizing on chance characteristics of the data (MacCallum et al., 1996).

Do Fit Indices for Heterosexual Men and Heterosexual Women Differ When They Evaluate Gay Men? The evaluation of the configural invariance model for heterosexual men and heterosexual women rating gay men showed an RMSEA statistic that was smaller than McDonald’s (1999) minimally recommended value with CFI and SRMR values considered having adequate fit as proposed by Hu and Bentler (1999; see Table 2). The configural model provided an adequate description for heterosexual men and heterosexual women rating gay men. This finding means that the one-factor structure provided equal patterns of factor loadings across groups. The weak invariance model was then assessed by constraining the factor loadings to equality across groups. Given the value of all three indices and the small differences in CFI and RMSEA, the weak invariant model provided an adequate description of the data. This finding means that factor loadings did not differ across groups. We then tested the strong invariance model by constraining factor loadings and latent intercepts to equality across groups. Table 2 shows that the RMSEA and SRMR did not provide adequate fit for the strong invariant model. After examining the modification indices and local fit information of the latent intercepts, model fit could be improved if three intercepts were allowed to differ across groups. The partial strong invariance model constrained factor loadings and latent intercepts; however, intercepts were not equal across groups. Heterosexual men compared to heterosexual women had greater intercepts for the following items:

1549

Invariance of the Modern Homonegativity Scale TABLE 2 Results of the invariance analyses as a function of rater gender Fit indices for heterosexual men and heterosexual women evaluating gay men Model

SB-χ 2

df

1 2 3 4 5

218.12 240.84 307.18 280.99 252.22

106 117 129 126 125

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Model Comparison

RMSEA (90% CI)

CFI

NNFI

SRMR

.074 .075 .085 .081 .073

.970 .966 .952 .958 .965

.962 .962 .950 .956 .964

.072 .099 .103 .098 .099

(.061, (.061, (.073, (.068, (.060,

.088) .088) .097) .093) .086)

ΔRMSEA

ΔCFI

ΔNNFI

ΔSRMR

−.001 −.01 .004

.004 .014 −.006

0 .012 −.006

−.027 −.004 .005

Model 1 vs. Model 2 Model 2 vs. Model 3 Model 3 vs. Model 4

Fit indices for heterosexual men and heterosexual women evaluating lesbians Model

SB-χ 2

df

1 2 3 4 5

235.63 243.98 310.51 277.81 254.63

106 117 129 126 125

RMSEA (90% CI)

CFI

NNFI

SRMR

.079 .074 .084 .078 .072

.963 .964 .949 .957 .963

.954 .960 .947 .955 .961

.073 .077 .101 .091 .076

(.065, (.061, (.072, (.065, (.060,

.092) .087) .096) .091) .085)

Model Comparison

ΔRMSEA

ΔCFI

ΔNNFI

ΔSRMR

Model 1 vs. Model 2 Model 2 vs. Model 3 Model 3 vs. Model 4

.005 −.01 .006

−.001 .015 −.008

−.006 .013 −.008

−.004 −.024 .01

Note: Model Model Model Model Model

1 2 3 4 5

refers to the configural invariant factor model. refers to the weak invariant factor model. refers to the strong invariant factor model. refers to the partial strong invariant factor model. allows the latent mean to be estimated.

Item 8, which reads, “If gay men want to be treated like everyone else, then they need to stop making such a fuss about their sexuality/culture”; and Item 9, which reads, “Gay men who are ‘out of the closet’ should be admired for their courage”—reverse coded. However, heterosexual women had a greater intercept than heterosexual men on Item 11, which reads “In today’s economic times, tax dollars shouldn’t be used to support gay male organizations.” The partial strong invariance model was evaluated and showed that RMSEA, CFI, and NNFI improved to indicate satisfactory model fit (relative to the weak invariance model) and demonstrated nine equal intercepts across groups (See Table 2).

1550

D. H. Romero et al.

In other words, the MHS had factor loadings that did not differ across groups, and most of the latent item intercepts did not differ across groups.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Do Fit Indices for Heterosexual Men and Heterosexual Women Differ When They Evaluate Lesbian Women? The evaluation of the configural invariance model for heterosexual men and heterosexual women rating lesbian women showed an RMSEA statistic that was smaller than minimally recommended (McDonald, 1999), and NNFI, CFI, and SRMR values provided adequate fit, which satisfied the proposed guidelines (Blackburn et al., 2004; Hu & Bentler, 1999; see Table 2). The configural model was considered acceptable and demonstrated the same pattern of factor loadings across groups. The weak invariance model had an improved RMSEA, CFI, and NNFI (see Table 2). Since the CFI’s differences between the configural and weak invariance models did not exceed .01, the weak invariant model provides an adequate description of the data, which indicated that the factor loadings did not differ across groups. The strong invariant model did not provide adequate fit where more than one latent intercept differed across groups (see Table 2). After examining the modification indices and local fit information, model fit could be improved if three intercepts were allowed to differ across groups. In rating lesbian women, heterosexual men compared to heterosexual women had higher intercepts for the following items: Items 5 reads, “Celebrations such as ‘Gay Pride Day’ are ridiculous because they assume that an individual’s sexual orientation should constitute a source of pride”; Item 11 reads, “In today’s tough economic times, tax dollars shouldn’t be used to support lesbian/gay male organizations”; and Item 12 reads, “Lesbians (gay men) have become far too confrontational in their demand for equal rights.” A partial strong invariant model that allowed these latent item intercepts to differ but required the other latent item intercepts to be identical across groups showed an improved RMSEA, SRMR, NNFI, and CFI, indicating good model fit (see Table 2).

Do Fit Indices for Heterosexual Men Differ When They Evaluate Gay Men and Lesbian Women? The evaluation of the configural invariance model for heterosexual men rating gay men and another group of heterosexual men rating lesbian women showed good model fit as evidenced by NNFI, CFI, and SRMR. RMSEA provided a minimally adequate fit (see Table 3). The configural invariance model was deemed acceptable and provided support for an equivalent pattern of factor loadings across groups. The weak invariance model had an RMSEA and CFI that provided acceptable model fit, and this model

1551

Invariance of the Modern Homonegativity Scale

TABLE 3 Results of the invariance analyses as a function of the sexual orientation of those rated Fit indices for heterosexual men evaluating gay men and lesbians Model

SB-χ 2

df

1 2 3 4 5

210.19 237.56 337.79 265.86 251.74

106 117 129 126 125

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Model Comparison

RMSEA (90% CI)

CFI

NNFI

SRMR

.078 .079 .096 .082 .079

.963 .957 .925 .950 .955

.954 .951 .924 .948 .952

.073 .099 .115 .102 .099

(.062, (.064, (.086, (.068, (.064,

.093) .094) .133) .096) .092)

ΔRMSEA

ΔCFI

ΔNNFI

ΔSRMR

−.001

.006

.003

−.026

−.003

.007

.003

−.003

Model 1 vs. Model 2 Model 2 vs. Model 3 Model 2 vs. Model 4

Fit indices for heterosexual women evaluating gay men and lesbians Model

SB-χ 2

df

1 2 3 4 5

243.13 250.52 333.89 289.84 268.35

106 117 129 127 126

RMSEA (90% CI)

CFI

NNFI

SRMR

.076 .072 .084 .076 .071

.969 .970 .954 .963 .968

.961 .966 .953 .962 .966

.065 .069 .106 .088 .069

(.064, (.059, (.073, (.064, (.059,

.089) .084) .096) .087) .083)

Model Comparison

ΔRMSEA

ΔCFI

ΔNNFI

ΔSRMR

Model 1 vs. Model 2 Model 2 vs. Model 3 Model 3 vs. Model 4

.004 −.012 .008

−.001 .016 −.009

−.005 .013 −.009

−.004 −.037 .018

Note. Model 1 refers to the configural invariant factor model. Model 2 refers to the weak invariant factor model. Model 3 refers to the strong invariant factor model. Model 4 refers to the partial strong invariant factor model. Model 5 allows the latent mean to be estimated.

demonstrated the factor loadings did not differ across groups. As evidence of this assertion, the CFI difference between the configural and weak invariant models did not exceed .01, and the weak invariance model had an acceptable NNFI (see Table 3). The estimation of the strong invariance model indicated poor model fit, as both the CFI and RMSEA indices deteriorated when compared to the weak invariance model. In fact, the intermediate solution is reported, as this model did not converge after 10,000 iterations. Modification fit indices of the latent intercepts were examined, and three intercepts were allowed to differ across groups. Heterosexual men who evaluated gay men had higher intercepts than another group of heterosexual men who evaluated lesbian women for Item 2, which reads, “Gay men/lesbians seem to focus on the ways in which they differ from heterosexuals, and ignore the ways in

1552

D. H. Romero et al.

which they are the same”; Item 7, which reads, “Gay men/lesbians should stop shoving their lifestyles down other people’s throats”; and Item 8, which reads, “If gay men/lesbians want to be treated like everyone else, then they need to stop making such a fuss about their sexuality/culture.” All fit indices were improved in the partial strong invariant model and indicated that nine of the intercepts did not differ across groups (see Table 3).

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Do Fit Indices for Heterosexual Women Differ When They Evaluate Gay Men and Lesbian Women? The evaluation of the configural invariance model for heterosexual women rating gay men and lesbian women showed an acceptable RMSEA, while NNFI, CFI, and SRMR provided adequate model fit (see Table 3). The configural model was deemed acceptable and demonstrated same patterns of factor loadings across groups. The weak invariance model improved in all but one fit index, which indicates adequate fit and demonstrates that the factor loadings did not differ across groups. Next, the strong invariant model was evaluated relative to the weak invariant model, and it did not provide adequate fit. After examining the modification indices of latent intercepts, model fit could be improved if two intercepts were allowed to differ across groups (see Table 3). Heterosexual women evaluating lesbian women started off at a higher intercept than heterosexual women evaluating gay men for the following items: Item 6 reads, “Gay men/lesbians still need to protest for equal rights”—reverse scored; and Item 9 reads, “Gay men/lesbians who are ‘out of the closet’ should be admired for their courage”—reverse scored. The evaluation of the partial strong invariant model showed improvements on all fit indices, indicating adequate fit and that 10 latent item intercepts did not differ across groups (see Table 3).

Group Comparisons at the Latent Mean Level As a partial strong invariance model held across all group comparisons, we then assessed latent mean differences across all group comparisons. Differences between heterosexual men and heterosexual women evaluating gay men were estimated at the latent mean level. The difference at the latent mean level across groups equaled −.31 (SE = .07, t = −4.47, p < .01, Cohen’s d = .53), indicating that heterosexual men compared to heterosexual women had greater negative attitudes toward gay men of more than one half a standard deviation. In addition, differences between heterosexual men and heterosexual women rating lesbian women estimated at the latent mean level equaled −.09 (SE = .04, t = −2.23, p < .05, Cohen’s d = .26), indicating that heterosexual men compared to heterosexual women had more negative attitudes toward lesbian women by one quarter of a standard deviation. Differences between

Invariance of the Modern Homonegativity Scale

1553

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

the heterosexual men rating gay men and the heterosexual men rating lesbian women estimated at the latent mean level across target groups equaled .02 (SE = .06, t = .42, p = ns, Cohen’s d = .05), indicating that heterosexual men had similar negative attitudes toward gay men and lesbian women. At the latent mean level, differences between heterosexual women evaluating lesbian women and another group of heterosexual women evaluating gay men were also estimated. The difference in the latent mean level across target groups was equal to .002 (SE = .04, t = .05, p = ns, Cohen’s d = .01), indicating that heterosexual women did not differ on their negative attitudes toward gay men and lesbian women.

DISCUSSION The current study demonstrated that group differences on the MHS were not an artifact of the measure but were due to “true” differences in homonegativity across participant sex. In particular, heterosexual men compared to heterosexual women held more negative attitudes toward both gay men and lesbian women. There were no differences in attitudes toward gay men and lesbian women when rated by either heterosexual men or heterosexual women.

Similarities to Past Research and Latent Mean Differences The pattern of study results was similar to that found by Morrison and Morrison (2003). First, the MHS tends to have acceptable internal scale score reliability in a sample of predominantly Mexican American college students. Second, in the evaluation of gay men, our findings demonstrated reduced homonegativity relative to past research. Heterosexual men from this study averaged 37.77 (SD = 9.25), compared to the previously reported heterosexual men’s average of 41.9 (SD = 9.1). Heterosexual women from this study averaged 32.93 (SD = 9.01) versus the previously reported heterosexual women’s average of 37.3 (SD = 10.1). Like Morrison and Morrison (2003), we also found that heterosexual men had greater latent mean homonegativity scores than women. In the evaluation of lesbian women, this study demonstrates reduced levels of homonegativity relative to the Morrison and Morrison (2003) findings. Heterosexual men from this study had an average of 36.31 (SD = 8.76), compared to the previously reported heterosexual men’s average of 42.8 (SD = 9.7). Heterosexual women from this study had an average of 33.48 (SD = 8.81), which compares to previously reported heterosexual women’s average of 38.8 (SD = 8.5). It is also notable that scores on a measure of social desirability were unassociated with modern homonegativity scores, suggesting that lower scores on the MHS were unlikely due to impression management.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

1554

D. H. Romero et al.

While these differences occur at the observed score level, the most important comparison takes place at the latent variable level. First, the one-factor model of the MHS was supported across all group comparisons, providing evidence of the unidimensionality of the MHS. Moreover, the amount of heterosexual male homonegativity toward gay men (d = 0.53) was more than twice the amount of heterosexual male homonegativity toward lesbian women (d = 0.26). One shortcoming of this type of analysis is that we cannot attribute latent mean differences to variables such as gender roles (Fagot, 1985; Kilianski, 2003) or masculinity (O’Neal, 1981). Nonetheless, our results indicate there are “true” differences in homonegativity as a function of rater sex. This finding of increased heterosexual male modern homonegativity toward lesbian women (relative to that of heterosexual women) contradicts earlier research that indicates that heterosexual men compared to heterosexual women would be less prejudiced toward lesbian women (Herek, 1988; Louderback & Whitley, 1997; Whitley, 1988). Some of this earlier research used the ATLG, which may explain the different results. Other variables such as the differing composition of the samples across studies may account for the different conclusions. For example, the emphasis of Mexican culture on clear and distinct gender roles may overwhelm other factors to which this difference has been attributed in earlier research, such as the eroticization of lesbian women by heterosexual men (Louderback & Whitley, 1997). At the latent mean level, there were no differences in modern homonegativity toward gay men and lesbian women when the raters were heterosexual women, which supports the previous findings of Morrison and Morrison (2003). However, there were also no significant differences between modern homonegativity toward gay men and lesbian women when the raters were heterosexual men. This finding contradicted our hypothesis that heterosexual men would express more homonegativity toward gay men (Herek, 1988; LaMar & Kite, 1998; VanderStoep & Green, 1988; Whitley, 1988). Again, the contradictory findings may be a confluence of sample composition and measure used to assess homonegativity. It is notable that heterosexual males did report some higher latent item intercepts of homonegativity toward gay men than toward lesbian women on the most judgmental items of the MHS. Using the ATLG, Herek (1988) also found that heterosexual men have greater levels of homonegativity toward gay men than toward lesbian women. Since items in the ATLG are known to be explicitly judgmental (Morrison & Morrison, 2003), this shows consistency in the way that heterosexual men respond on such items. This response pattern may reflect a “protective sexism” in which heterosexual males are more comfortable in endorsing confrontational language toward their male peers who identify as gay than they are in doing so toward females who identify as lesbian. This may be due to concerns of heteronormative men being viewed as feminine or insufficiently masculine and, as a result, may demonstrate or confirm their masculinity by being harsh

Invariance of the Modern Homonegativity Scale

1555

and antagonistic toward gay men (Theodore & Basow, 2000; Wilkinson, 2004). It is important to acknowledge these findings, since heterosexual men may continue to have high levels of modern homonegativity when endorsing judgmental items.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Limitations and Future Directions Several methodological limitations of the current study should be addressed. First, we used a highly simplified measure of sexual orientation, based solely on self-identification and including only three categorical options (homosexuality, heterosexuality, or bisexuality). A continuous scale may have revealed greater variability in orientation (Drucker, 2010), and dichotomizations and other truncations of continuous variables are known to result in attenuated effect sizes and reduced statistical power (Cohen, 1983; MacCallum, Zhang, Preacher, & Rucker, 2002). However, the purpose of the measure in the present study was for the inclusion or exclusion of participants, a purpose for which a dichotomous decision is necessary. In addition, we cannot rule out the possibility of an order effect, since the ATLG was administered to all participants before the MHS, potentially affecting the scores on the MHS (Podsakoff et al., 2003). However, we have no reason to believe that an order effect would differentially impact results across groups. The predominantly Latino college student sample in the present study is of particular interest. In Latino culture, attitudes toward gay men are often complex. Gay Latinos may be perceived as lacking culturally sanctioned “machismo” and often endure stigma that results in psychological distress (Díaz, 1998; Díaz, Ayala, Bein, Henne, & Marín, 2001). As noted above, our male participants did show slightly greater endorsement of the most judgmental items on the MHS when rating gay men. However, the relative youth of our sample may temper expressed homonegativity. Future research should examine changes in attitudes over time across cultures that have traditionally differed in expressed homonegativity. Additional work should focus on why some male heterosexual college students continue to endorse items that are confrontational in the face of research demonstrating lower homonegativity among teens (Savin-Williams, 2005). In summary, the MHS demonstrates full invariance of the factor loadings and partial invariance of latent item intercepts across the sex of raters and rated targets. The measure can appropriately be used across these groups to study homonegativity, and the current research extends the findings of Morrison and Morrison (2003) to highlight “true” gender differences in modern homonegativity, such that heterosexual men (relative to heterosexual women) express increased homonegativity toward both gay men and lesbian women.

1556

D. H. Romero et al.

ACKNOWLEDGMENT This research was based on the Master’s thesis completed by the first author while he was a student at the University of Texas at El Paso.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

REFERENCES Aberson, C. L., Swan, D. J., & Emerson, E. P. (1999). Covert discrimination against gay men by U.S. college students. Journal of Social Psychology, 139, 323–334. doi:10.1080/00224549909598388 Barron, J. M., Struckman-Johnson, C., Quevillon, R., & Banka, S. R. (2008). Heterosexual men’s attitudes toward gay men: A hierarchical model including masculinity, openness, and theoretical explanations. Psychology of Men & Masculinity, 9, 154–166. doi:10.1037/1524-9220.9.3.154 Blackburn, R., Donnelly, J. P., Logan, C., & Renwick, S. J. D. (2004). Convergent and discriminative validity of interview and questionnaire measures of personality disorder in mentally disordered offenders: A multitrait–multimethod analysis using confirmatory factor analysis. Journal of Personality Disorders, 18, 129–150. doi:10.1521/pedi.18.2.129.32779 Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY: Guilford. Cárdenas, M., & Eduardo Barrientos, J. (2008). The Attitudes Toward Lesbians and Gay Men Scale (ATLG): Adaptation and testing the reliability and validity in Chile. Journal of Sex Research, 45, 140–149. doi:10.1080/00224490801987424 Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14, 464–504. doi:10.1080/ 10705510701301834 Cheung, G. W., & Rensvold, R. B. (2001). The effects of model parsimony and sampling error on the fit of structural equation models. Organizational Research Methods, 4, 236–264. doi:10.1177/109442810143004 Cohen, J. (1983). The cost of dichotomization. Applied Psychological Measurement, 7, 249–253. doi:10.1177/014662168300700301 Cotton-Huston, A. L., & Waite, B. M. (2000). Anti-homosexual attitudes in college students: Predictors and classroom interventions. Journal of Homosexuality, 38, 256–268. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349–354. doi:10.1037/h0047358 Díaz, R. M. (1998). Latino gay men and HIV: Culture, sexuality and risk. New York, NY: Routledge. Díaz, R. M., Ayala, G., Bein, E., Henne, J., & Marín, B. (2001). The impact of homophobia, poverty and racism on the mental health of gay and bisexual Latino men: Findings from 3 U.S. cities. American Journal of Public Health, 91, 927–932. doi:10.2105/AJPH.91.6.927

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Invariance of the Modern Homonegativity Scale

1557

Drucker, D. J. (2010). Male sexuality and Alfred Kinsey’s 0-6 scale: Toward a “sound understanding of the realities of sex.” Journal of Homosexuality, 57, 1105–1123. doi:10.1080/00918369.2010.508314 Enders, C. K. (2010). Applied missing data analysis. New York, NY: Guilford. Fagot, B. I. (1985). Beyond the reinforcement principle: Another step toward understanding sex role development. Developmental Psychology, 21, 1097–1104. doi:10.1037/0012-1649.21.6.1097 Herek, G. M. (1984a). Attitudes toward lesbians and gay men: A factor analytic study. Journal of Homosexuality, 10, 39–51. doi:10.1300/J082v10n01_03 Herek, G. M. (1984b). Beyond “Homophobia”: A social psychological perspective on attitudes toward lesbians and gay men. Journal of Homosexuality, 10, 1–21. doi:10.1300/J082v10n01_01 Herek, G. M. (1988). Heterosexuals’ attitudes toward lesbians and gay men: Correlates and gender differences. Journal of Sex Research, 25, 451–477. doi:10.1080/ 00224498809551476 Horn, J. L., McArdle, J. J., & Mason, R. (1983). When is invariance not invariant: A practical scientist’s look at the ethereal concept of factor invariance. Southern Psychologist, 1, 179–188. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55. doi:10.1080/ 10705519909540118 Jöreskog, K. G., & Sörbom, D. (2003). PRELIS 2: User’s reference guide: A program for multivariate data screening and data summarization: A preprocessor for LISREL. Lincolnwood, IL: Scientific Software International. Jöreskog, K. G., & Sörbom, D. (2006). Lisrel (version 8.80) [Computer Software]. Lincolnwood, IL: Scientific Software International. Kilianski, S. E. (2003). Explaining heterosexual men’s attitudes toward women and gay men: The theory of exclusively masculine identity. Psychology of Men & Masculinity, 4, 37–56. doi:10.1037/1524-9220.4.1.37 Kite, M. E., & Deaux, K. (1987). Gender belief systems: Homosexuality and the implicit inversion theory. Psychology of Women Quarterly, 11, 83–96. doi:10.1111/j.1471-6402.1987.tb00776.x Kite, M. E., & Whitley, Jr., B. E. (1996). Sex differences in attitudes toward homosexual persons, behaviors and civil rights: A meta-analysis. Personality and Social Psychology Bulletin, 22, 336–353. doi:10.1177/0146167296224002 Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford. LaMar, L., & Kite, M. (1998). Sex differences in attitudes toward gay men and lesbians: A multidimensional perspective. Journal of Sex Research, 35, 189–196. doi:10.1080/00224499809551932 Lehavot, K., & Lambert, A. J. (2007). Toward a greater understanding of antigay prejudice: On the role of sexual orientation and gender role violation. Basic and Applied Social Psychology, 29, 279–292. doi:10.1080/01973530701503390 Louderback, L. A., & Whitley, Jr., B. E. (1997). Perceived erotic value of homosexuality and sex-role attitudes as mediators of sex differences in heterosexual

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

1558

D. H. Romero et al.

college students’ attitudes toward lesbians and gay men. Journal of Sex Research, 34, 175–182. doi:10.1080/00224499709551882 MacCallum, R. C. (2003). Working with imperfect models. Multivariate Behavioral Research, 38, 113–139. doi:10.1207/S15327906MBR3801_5 MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods, 1, 130–149. doi:10.1037/1082-989X.1.2.130 MacCallum, R. C., Zhang, S., Preacher, K. J., & Rucker, D. D. (2002). On the practice of dichotomization of quantitative variables. Psychological Methods, 7, 19–40. doi:10.1037/1082-989X.7.1.19 Marsiglio, W. (1993). Attitudes toward homosexual activity and gays as friends: A national survey of heterosexual 15- to 19-year old males. Journal of Sex Research, 30, 12–17. doi:10.1080/00224499309551673 McCreary, D. R. (1994). The male role and avoiding femininity. Sex Roles, 31, 517–531. doi:10.1007/BF01544277 McDonald, R. P. (1999). Test theory: A unified treatment. Mahwah, NJ: Erlbaum. Meade, A. W., & Wright, N. A. (2012). Solving the measurement invariance anchor item problem in item response theory. Journal of Applied Psychology, 97, 1016–1031. doi:10.1037/a0027934 Mohipp, C., & Morry, M. M. (2004). The relationship of symbolic beliefs and prior contact to heterosexuals’ attitudes toward gay men and lesbian women. Canadian Journal of Behavioral Science, 36, 36–44. doi:10.1037/h0087214 Morrison, M. A., & Morrison, T. G. (2003). Development and validation of a scale measuring modern prejudice toward gay men and lesbian women. Journal of Homosexuality, 43, 15–37. doi:10.1300/J082v43n02_02 Morrison, M. A., & Morrison, T. G. (2011). Sexual orientation bias toward gay men and lesbian women: Modern homonegative attitudes and their association with discriminatory behavioral intentions. Journal of Applied Social Psychology, 41, 2573–2599. doi:10.1111/jasp.2011.41.issue-11 Morrison, M. A., Morrison, T. G., & Franklin, R. (2009). Modern and old-fashioned homonegativity among samples of Canadian and American university students. Journal of Cross-Cultural Psychology, 40, 523–542. doi:10.1177/ 0022022109335053 O’Neal, J. M. (1981). Male sex role conflicts, sexism, and masculinity: Psychological implications for men, women, and the counseling psychologist. Counseling Psychologist, 9, 61–80. doi:10.1177/001100008100900213 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(5), 879–903. doi:10.1037/0021-9010.88.5.879 Preacher, K. J., & Coffman, D. L. (2006, May). Computing power and minimum sample size for RMSEA [Computer software]. Retrieved from http://www. quantpsy.org Ramirez, M., & Carrasco, N. (1996). Revision of the family attitude scale. Unpublished manuscript.

Downloaded by [University of Lethbridge] at 21:05 28 September 2015

Invariance of the Modern Homonegativity Scale

1559

Rye, B. J., & Meaney, G. J. (2010). Measuring homonegativity: A psychometric analysis. Canadian Journal of Behavioral Science, 42, 158–167. doi:10.1037/ a0018237 Satcher, J., & Legget, M. (2006). Homonegativity among Alabama counselors. Alabama Counseling Association Journal, 32, 1–11. Satcher, J., & Legget, M. (2007). Homonegativity among professional school counselors: An exploratory study. Professional School Counseling Journal, 11, 10–16. doi:10.5330/PSC.n.2010-11.10 Satorra, A., & Bentler, P. M. (1994). Corrections to test statistics and standard errors in covariance structure analysis. In A. von Eye, & C. C. Clogg (Eds.), Latent variable analysis: Applications for developmental research (pp. 399–419). Thousand Oaks, CA: Sage Publications. Savin-Williams, R. C. (2005). The new gay teenager. Cambridge, MA: Harvard University Press. Schafer, J. L. (1999). NORM: Multiple imputation of incomplete multivariate data under a normal model, version 2. Software for Windows 95/98/NT. Retrieved from http://www.stat.psu.edu/~jls/misoftwa.html Schmitt, N., & Kuljanin, G. (2008). Measurement invariance: Review of practice and implications. Human Resource Management Review, 18, 210–222. doi:10.1016/j. hrmr.2008.03.003 Steenkamp, J.-B., & Baumgartner, H. (1998). Assessing measurement invariance in cross national consumer research. Journal of Consumer Research, 25, 78–107. doi:10.1086/209528 Theodore, P. S., & Basow, S. A. (2000). Heterosexual masculinity and homophobia: A reaction to the self? Journal of Homosexuality, 40, 31–48. doi:10.1300/ J082v40n02_03 Thissen, D., Steinberg, L., & Gerrard, M. (1986). Beyond group mean differences: The concept of item bias. Psychological Bulletin, 99, 118–128. doi:10.1037/00332909.99.1.118 VanderStoep, S. W., & Green, C. W. (1988). Religiosity and homonegativism: A pathanalytic study. Basic and Applied Social Psychology, 9, 135–147. doi:10.1207/ s15324834basp0902_5 Whitley, B. E. (1988). Sex differences in heterosexuals’ attitudes toward homosexuals: It depends upon what you ask. Journal of Sex Research, 24, 287–291. doi:10.1080/00224498809551426 Wilkinson, W. W. (2004). Authoritarian hegemony, dimensions of masculinity, and male antigay attitudes. Psychology of Men & Masculinity, 5, 121–131. doi:10.1037/1524-9220.5.2.121

Assessing the Gender Invariance of the Modern Homonegativity Scale.

The measurement invariance of the Modern Homonegativity Scale (MHS) was examined among heterosexual female (n = 449) and male (n = 329) university stu...
347KB Sizes 0 Downloads 6 Views