Multiple-group confirmatory factor analysis has been used to test measurement equivalence across groups (Cheung & Rensvold, 1999; Vandenberg & Lance, 2000). Noninvariance suggests that a measure does not reliably measure the same construct in both groups. Several levels of measurement invariance have been suggested (Bontempo, Hofer, & Lawrence, 2006; Vandenberg & Lance, 2000), but at the very least should include equivalence in factor loadings across groups. Within the field of prevention science, Gottfredson and colleagues (Gottfredson & Koper, 1996, 1997; Rosay, Gottfredson, Armstrong, & Harmon, 2000) have conducted a number of studies using latent variable structural techniques to assess measurement invariance as well as predictive relationships. In the first two studies (Gottfredson & Koper, 1996, 1997), results supported equivalence of nearly all of the measures of risk factors for drug use (e.g., school attachment, problem behaviors) between African American and European American adolescents. In a third study (Rosay, Gottfredson, Armstrong, & Harmon, 2000), the authors focused on the assessment of invariance of measures of prevention program effectiveness across five racial/ethnic groups (African American, Hispanic American, Asian American, American Indian, and
