Fit, evaluate, and visualise generalised additive models (GAMs) in native Julia
Generalised additive models (GAMs) are an incredibly powerful modelling tool for regression practitioners. However, the functionality of the excellent R package mgcv
is yet to be built in native Julia. This package aims to do just that, albeit at much less complexity given how sophisticated mgcv
is.
GAM.jl
is very much a work in progress. Please check back for updates and new features!
Currently, GAM.jl
fits the following model:
where
The smooth term
where
The spline basis functions are defined as:
$$ B_{k,j}(x_j) = \prod_{m=1}^{d} (x_j - \text{knots}{k,m})^{[x_j \geq \text{knots}{k,m}]} , $$
where
The spline basis functions are defined such that they are zero outside of the range of the knots. The coefficients
where
where