I'm following up on this great answer regarding running Principal Component Analysis (PCA) to uncover the reason behind lack of convergence and/or singularity for Mixed-Effects models.
My model below doesn't convergence, however, I wonder why I don't get a convergence when model is not singular?
NOTE: When we drop the correlation between intercepts and slope the model below becomes singular! i.e.,
lmer(math ~ ses*sector + (ses || sch.id), data = dat)
library(lme4) dat <- read.csv('https://raw.githubusercontent.com/rnorouzian/e/master/nc.csv') m4 <- lmer(math ~ ses*sector + (ses | sch.id), data = dat) summary(m4) Random effects: Groups Name Variance Std.Dev. Corr sch.id (Intercept) 3.6302 1.9053 ses 0.7356 0.8577 0.46 Residual 10.1951 3.1930 Number of obs: 7185, groups: sch.id, 160 summary(rePCA(m4)) $sch.id Importance of components: [,1] [,2] Standard deviation 0.6118 0.2321 Proportion of Variance 0.8742 0.1258 Cumulative Proportion 0.8742 1.0000