Parameters that were sampled in cells from multiple microscopic fields and from multiple cell lines were fit to a linear mixed-effects model using the lme4 package in R [34]. The model included the cell line identifier and sex. A Tukey post-hoc test identified pairwise differences in two-factor models (e.g., genotype and ethanol). Electrophysiology results were modeled using generalized estimating equations (GEE) with the cell line as grouping identifier using the geepack R package [35]. Where appropriate, ANOVA or Student’s t test was used as indicated in figure legends. For RNAseq, pseudo-bulk data (single-cell reads pooled by subject identifier) were modeled in DESeq2 [36], testing first by likelihood ratio testing (LRT) over all groups, and then pairwise comparisons were evaluated using Wald tests. After Benjamini-Hochberg multiple measurements correction, a false discovery rate of 5% was set as a threshold for significance. Numbers of replicate cells and/or fields per cell line and genotype used for statistical testing are listed in Supplementary Table 6.
