Recent developments have enabled the modeling of longitudinal assessment data in a diagnostic classification model (DCM) framework. These longitudinal DCMs were developed to provide measures of student growth on a discrete scale in the form of attribute mastery transitions, thereby supporting categorical and criterion-referenced interpretations of growth. Studies employing longitudinal DCMs have used different statistical approaches to model examinee attribute mastery transitions. Yet, there has not been research that systematically compares the potential advantages and shortcomings of these different approaches. Via simulation, this study compares and evaluates the performance of three different approaches to estimating longitudinal DCMs. Results show that performance is similar in terms of classification accuracy and reliability, but practical considerations and the overall goals of the application should guide the choice of modeling approach. Implications of these results are discussed. [This is the online version of an article published in "Behaviormetrika."]