Support for linear mixed effects models

[grst]

Description of feature

They can be used to either perform tests on all cells, including the sample as random effect, or on pseudobulk when there are designs that require random effects.

Methods

• MAST (add MAST (via rpy2) #19)
• DREAM
• Statsmodels LME

Model specification

• Neither formulaic nor patsy currently allow for random effects. There's a PR open to add support Draft: Add support for mixed effects. matthewwardrop/formulaic#34
• formulae allows specifying random effects: https://github.com/bambinos/formulae
• Alternatively, it could be possible to specify two formulas, one for fixed and one for random effects.

[grst]

Migrated to scverse/pertpy#614