Metida
Mixed Models
Metida.jl is a Julia package for fitting mixed-effects models with flexible covariance structure. At this moment package is in early development stage.
Main goal to make reproducible output corresponding to SAS/SPSS.
Now implemented covariance structures:
- Scaled Identity (SI)
- Variance Components / Diagonal (VC)
- Autoregressive (AR)
- Heterogeneous Autoregressive (ARH)
- Compound Symmetry (CS)
- Heterogeneous Compound Symmetry (CSH)
Usage:
LMM(model, data; subject = nothing, random = nothing, repeated = nothing)
where
model
is a fixed-effect model (@formula
), example: @formula(var ~ sequence + period + formulation)
random
vector of random effects or single random effect. Effect can be declared like this: VarEffect(@covstr(formulation), CSH)
. @covstr
is a effect model: @covstr(formulation)
. CSH
is a CovarianceType structure. Premade constants: SI, VC, AR, ARH, CSH.
repeated
is a repeated effect (only single).
subject
is a block-diagonal factor.
See also:
https://github.com/JuliaStats/MixedModels.jl