Julia wrapper for the alpaca R library to estimate generalized linear model with high-dimensional fixed effects.
Alpaca.jl is currently targeting alpaca 0.3.3.
You need to have R, and Julia of course, preinstalled, but if either is 64-bit the other needs to match (so 32-bit R and 64-bit Julia are incompatible):
] add Alpaca
and in R:
> install.packages('alpaca')
> install.packages('formula.tools')
Use the feglm function. See the following example:
using Alpaca, RDatasets, Distributions, Random
# setting up the example data
rng = MersenneTwister(1234)
df = dataset("datasets", "iris")
df.binary = vec(Float64.(rand(rng,0:1,size(df,1),1)))
df[!,:SpeciesDummy] = categorical(df[!,:Species])
idx = rand(rng,1:3,size(df,1),1)
a = ["A","B","C"]
df.Random = vec([a[i] for i in idx])
df.RandomCategorical = categorical(df[!,:Random])
# estimating the model
result = feglm(df, @formula(binary ~ SepalWidth), Binomial(),
fe = :(SpeciesDummy + RandomCategorical),
vcov = :(cluster(SpeciesDummy + RandomCategorical))
)The full form of the feglm function is
function feglm(df::AbstractDataFrame, f::FormulaTerm,
family::UnivariateDistribution;
fe::Union{Symbol, Expr, Nothing} = nothing,
vcov::Union{Symbol, Expr, Nothing} = :(simple()),
start::Union{Vector{T}, Nothing} = nothing,
maxiter::Integer = 10000, limit::Integer = 10,
trace::Integer = 0,
convtol::Real = 1.0e-06,
devtol::Real = 1.0e-08, steptol::Real = 1.0e-08,
centertol::Real = 1.0e-05, rhotol::Real = 1.0e-04,
droppc::Bool = true
) where T<:RealFor an explanation of the options, see the manual of the alpaca package.