Latent class counts for Christian nationalism items are unstable across samples

Evidence

In addition, we recognize that both Davis and Smith and Adler (2022) conduct latent class analysis (LCA) that treat the six CN items as ordinal variables in the BRS 2017 sample. Relying on the Bayesian Information Criterion (BIC) statistics, both studies suggest there should be ideally six latent groups who answer the six CN items distinctly. However, we find the modeling methodology is very sensitive to samples and measurements. We tried to replicate Davis’ and Smith and Adler’s findings in other BRS and ATP samples. BIC scores suggest either 6, 7, or 8 groups could be statistically ideal. The detailed results of LCA are presented in Table A5. While LCA is a powerful tool in recent research on American popular nationalisms, we think more robustness tests are necessary before one applies it to research using the six CN items. — Ruiqian Li and Paul Froese, The Duality of American Christian…, 92

Categorization

• XREF: Relevant to the reading purpose: LCA’s instability across samples is a methodological caution that hints at the heterogeneity of CN — supporting the authors’ factorial split rather than a unitary proxy for conservatism. — connects to: Christian nationalism splits into two ideals measured on acceptance spectrums rather than discrete classes — Ruiqian Li and Paul Froese, The Duality of American Christian…, 94

Broader Connections

• Christian nationalism splits into two ideals measured on acceptance spectrums rather than discrete classes #qualifies (0.80) — Ruiqian Li and Paul Froese, The Duality of American Christian…, 94