Hi,
I'm interested in how to implement calculation of partial derivatives/marginal effects similar to the package marginaleffects in r.
I translated some code from the excellent documentation, to simply calculate the marginal effect of hp for every unit in the dataset.
using GLM, RDatasets, DataFrames
mtcars = dataset("datasets", "mtcars")
rename!(mtcars, lowercase.(names(mtcars)))
lm_mod = glm(@formula(mpg ~ hp * cyl), mtcars, Normal())
partial_derivative = function(data, model, eps)
d1 = transform(data, :hp => (x -> x .- eps / 2) => :hp)
d2 = transform(data, :hp => (x -> x .+ eps / 2) => :hp)
(p1, p2) = map(d -> predict(model, d), [d1, d2])
round.((p2-p1)/eps, digits=2)
end
partial_derivative(mtcars, lm_mod, 1e-4)
What would be a reasonable way to implemented this in a way that could give me unit-level marginaleffects for every variable in the model efficiently?
I have been looking at ForwardDiff, but haven't been able to figure out how to do it.
Not sure I understand your question - are you just asking for how to do the partial_derivative
function for every variable in the model? That would just be partial_derivative(data, model, eps, var)
end than replace var
with whatever variable you're interested in, or use the StatsModels
API to extract all RHS variables and vall that function on all of them.
Not so much asking for ways to expand this function. But more so interested if there's some alternative implementation that would be faster/scale better with more variables and data.
I was looking through the code for Effects.jl but couldn't really understand the implementation they are using there.
Thanks for answering Nils, let's continue on the discourse (i xposted) there: https://discourse.julialang.org/t/how-to-implement-r-like-marginaleffects/114260
Last updated: Nov 22 2024 at 04:41 UTC