“…Therefore, it is urgent to establish theoretically-guaranteed, efficient, and easy-to-implement methods for LCM-based latent class analysis, along with criteria to evaluate algorithm performance in categorical data. On the other hand, spectral clustering algorithms based on eigen-decomposition or singular value decomposition (SVD) of certain matrices are popular techniques in machine learning, pattern recognition, statistical learning, and social network analysis (see (Ng et al, 2001;Von Luxburg, 2007;Rohe et al, 2011;Qin & Rohe, 2013;Lei & Rinaldo, 2015;Jin, 2015;Rohe et al, 2016;Binkiewicz et al, 2017;Mao et al, 2018Mao et al, , 2021Jin et al, 2023) and references therein) for their good theoretical properties, ease of implementation, and computational efficiency. However, they are rarely used for the problem of latent class analysis.…”