Latent classes of aggression and peer victimization: Measurement invariance and differential item functioning across sex, race-ethnicity, cohort, and study site

Abstract

Peer victimization is common and linked to maladjustment. Prior research has typically identified four peer victimization subgroups: aggressors, victims, aggressive-victims, and uninvolved. However, findings related to sex and racial-ethnic differences in subgroup membership have been mixed. Using data collected in September of 2002 and 2003, this study conducted confirmatory latent class analysis of a racially-ethnically diverse sample of 5415 sixth graders (49% boys; 50.6% Black; 20.9% Hispanic) representing two cohorts from 37 schools in four U.S. communities to replicate the four subgroups and evaluate measurement invariance of latent class indicators across cohort, sex, race-ethnicity, and study site. Results replicated the four-class solution and illustrated that sociodemographic differences in subgroup membership were less evident after accounting for differential item functioning.

FIGURE 1

Latent class analysis path diagrams of: (a) unconditional latent class measurement model; (b) MIMIC model with measurement invariance; (c) MIMIC model with uniform DIF; and (d) MIMIC model with nonuniform DIF. DIF, differential item functioning; MIMIC, multiple‐indicator‐multiple‐causes

FIGURE 2

Model‐estimated item response probability plot based on 4‐class confirmatory latent class analysis with boundary constraints indicated by horizontal lines (all uninvolved class items probabilities <.30 and all aggressive‐victims class item probabilities >.70) and inequality constraints indicated by marker size (non‐victimized aggressor class aggression‐related item probabilities greater than predominantly victimized class; non‐victimized aggressor class victimization‐related item probabilities less than predominantly victimized class)

FIGURE 3

Model‐estimated item response probability plots based on 4‐class multiple‐indicator‐multiple‐causes latent class analysis with sex, race‐ethnicity, site, and cohort as predictors of class membership and with sex, race‐ethnicity, and site as sources of DIF. Item responses that make up latent classes are listed along the x‐axis. Values shown on the y‐axis indicate the estimated probability of endorsing a particular item for participants within each latent class. Lines without markers indicate classes estimated at the mean of all covariates; markers indicate classes estimated for (i) Black boys in Georgia and (ii) White girls in Illinois to illustrate the magnitude of measurement non‐invariance from the combined DIF effects of sex, race‐ethnicity, and site. DIF, differential item functioning

FIGURE 4

Model‐estimated probabilities of class membership for boys versus girls (top) and Black versus Hispanic versus White students (bottom) based on 4‐class multiple‐indicator‐multiple‐causes latent class analysis with sex, race‐ethnicity, site, and cohort as predictors of class membership and with sex, race‐ethnicity, and site as sources of DIF (left) and 4‐class latent class regression model with no DIF effects (right). All values are adjusted for other covariates included in the model. Class 1 = uninvolved, Class 2 = aggressive‐victims, Class 3 = non‐victimized aggressors, and Class 4 = predominantly victimized. DIF, differential item functioning