How openpharma/mmrm works
mmrm is an R package for fitting marginal mixed models for repeated-measures continuous outcomes. The main workflow is to supply a formula and longitudinal data to the fitting interface, estimate fixed effects and covariance parameters with ML or REML, then summarize, test contrasts, or obtain estimated means and predictions. Covariance construction, cached factorizations and derivatives, and C++ likelihood calculations form the computational core. The README documents additional inference and ecosystem capabilities; sampled source excerpts establish the C++ component relationships, but do not expose all R-level dispatch wiring.
Main entry points: mmrm-methods.R Fit methods · fit.R Fit interface · mmrm.cpp Likelihood fitting · chol_cache.h Cholesky caches
Components 23
Model fitting
- Fit interface
R/fit.R - Formula utilities
R/utils-formula.R - Covariance specification
R/cov_struct.R - Fit methods
R/mmrm-methods.R
Numerical backend
- Covariance matrices
src/covariance.h - Cholesky caches
src/chol_cache.h - Covariance derivatives
src/derivatives.h - Likelihood fitting
src/mmrm.cpp - TMB interface
R/tmb.R - KR components
src/kr_comp.cpp
Inference and outputs
- Empirical covariance
R/empirical.R - Contrast testing
R/testing.R - Satterthwaite inference
R/satterthwaite.R - Kenward-Roger inference
R/kenwardroger.R - Estimated means
R/interop-emmeans.R - Prediction
src/predict.cpp - Residual diagnostics
R/residual.R - Tidy outputs
R/tidiers.R
Ecosystem integrations
- car integration
R/interop-car.R - parsnip engine
R/interop-parsnip.R - emmeans support
R/interop-emmeans.R
Other components
- Analyst
- Model consumer
Connections 24
- Analyst → Fit interface: supplies model
README.md - Fit interface → Formula utilities: uses formula
- Fit interface → Covariance specification: specifies covariance
- Fit interface → TMB interface: fits through
- TMB interface → Likelihood fitting: evaluates likelihood
- Likelihood fitting → Cholesky caches: gets factors
src/mmrm.cpp - Likelihood fitting → Covariance matrices: uses covariance
src/mmrm.cpp - Cholesky caches → Covariance matrices: builds factors
src/chol_cache.h - Fit interface → Fit methods: returns fit
- Fit methods → Satterthwaite inference: summarizes with
- Fit methods → Empirical covariance: supports robust inference
- Fit methods → Contrast testing: tests contrasts
- Contrast testing → Kenward-Roger inference: adjusts degrees freedom
- Kenward-Roger inference → KR components: computes components
- KR components → Covariance derivatives: gets derivatives
- Covariance derivatives → Cholesky caches: extends cache
src/derivatives.h - Fit methods → Estimated means: supports means
README.md - Fit methods → Prediction: predicts outcomes
- Fit methods → Residual diagnostics: extracts residuals
- Fit methods → Tidy outputs: formats results
- Fit methods → car integration: integrates with
README.md - Fit methods → parsnip engine: registers engine
README.md - Estimated means → emmeans support: uses support
README.md - Model consumer → Fit methods: consumes results