It is argued that studies of early adolescent peer victimization and bullying should be longitudinal because of the dynamism of this developmental period. Traditional methods for analyzing longitudinal data are inadequate because of the strict data requirements and inflexibility of models. A better alternative is linear mixed models (LMMs) for repeated measures. LMMs have less restrictive data requirements and much flexibility in the type of models that may be specified. Using empirical data from middle school students, it is shown how LMMs can be used to examine three major aspects of change in dominance and bullying. The first is unconditional change, which involves treating the sample as an entire group and modeling the mean trajectories of dominance and bullying over time. The second is conditional change, which involves examining gender differences in mean change of dominance and bullying over time. The third is dynamic change, which involves examining the longitudinal covariation between bullying and dominance controlling for trajectory effects. The algebra of the LMMs is presented from a multilevel perspective assuming a random effects model. The results are discussed in terms of dominance theory and highlight the advantages of the LMM approach to data analysis in the study of adolescent peer victimization and bullying.