lmmix fits Gaussian linear mixed models by maximum likelihood or REML. Its main use case combines random effects with structured residual covariance.
GitHub development version: 0.4.0. Not distributed on CRAN.
Models and inference
The package supports independent, compound-symmetric, AR(1), Toeplitz and unstructured residual covariance. It includes Satterthwaite and Kenward-Roger inference, model comparisons and estimated marginal means.
Scope and validation
The validation vignette documents comparisons with independent R implementations and PROC MIXED examples, including tolerances and limitations.
The current scope is univariate Gaussian responses. Some covariance matrices and inference calculations remain dense, so large-scale computation depends on the model structure.
Public sources checked on September 23, 2026.