Replication using confirmatory latent class analysis: A school climate example

The simultaneous growth in the availability of large publicly available datasets and the accessibility of powerful computation software to analyze those data have accelerated the use of mixture modeling techniques, including latent class analysis (LCA). Though secondary data analysis using LCA is increasingly common in school psychologyrelated research, there exist limits to the impacts of these analytic techniques. It is common to hear critiques that modeled results may be artifacts of a specific dataset rather than more stable and meaningful participant response patterns with practical implications. LCA-specific replication guidance is necessary to support an expansion of studies of this type. Via a school climate-related case example, we provide guidance for those wishing to shift from exploratory LCA to confirmatory LCA (CLCA). We scrutinized an earlier LCA analysis that identified a 5-class solution for student school climate survey data collected during the 2016-2017 academic year. Using student school climate perception data (n = 24,051) collected during the 2017–2018 academic year, our CLCA results suggest substantial similarity in response patterns. The initial exploratory results and this study's confirmatory results demonstrate five predominant patterns of student school climate experiences, suggesting that the 5-class solution replicates across data collection years and with entirely new respondents. Recognizing that CLCA methods are relatively nascent, we describe decision-making points researchers interested in utilizing CLCA may encounter and provide guidance to help avoid accusations of questionable research practices.