[stemangiola]

Summary

• Keep the numeric library-size offset on the formula environment so lme4 2.1 model.frame() no longer resolves offset to stats::offset().
• Stop with the first gene-level error when every glmmSeq fit fails, instead of returning that text and crashing the vignette on @stats.
• This unblocks the introduction vignette mixed-model chunk that is failing rworkflows on master.

Test plan

• Confirm rworkflows rebuilds introduction.Rmd, including the glmmseq_lme4 chunk
• Locally, with lme4 2.1, test_differential_expression(~ dex + (1 | cell), method = "glmmseq_lme4", formula_dispersion = ~ dex + cell) on a few airway genes returns finite p-values

Made with Cursor