Multiple-Group Analysis of Similarity in Latent Profile Solutions
提出一个六步流程，系统评估潜在剖面在配置、结构、离散、分布、预测和解释六个方面的相似性，并用北美和法国的组织承诺心态数据示例，帮助研究者判断剖面解在不同群体或时间点是否可推广。
Despite the increased popularity of person-centered analyses, no comprehensive approach exists to guide the systematic investigation of the similarity (or generalizability) of latent profiles, their predictors, and their outcomes across subgroups of participants or time points. We propose a six-step process to assess configural (number of profiles), structural (within-profile means), dispersion (within-profile variability), distributional (size of the profiles), predictive (relations between predictors and profile membership), and explanatory (relations between profile membership and outcomes) similarity. We then apply this approach to data on organizational commitment mindsets collected in North America (n = 492) and France (n = 476). This approach provides a rigorous method to systematically and quantitatively assess the extent to which a latent profile solution generalizes across diverse samples, such as in the cross-national comparison in our illustrative example, or the extent to which interventions or naturalistic changes may impact the nature of a latent profile solution. This approach also helps to identify the nature of any differences that might be present, thus providing richer interpretations of observed differences and ideas for future research.