Calculate marginal effects, similar to linear coefficients, for any predictive model. Also calculates and plots individual conditional expectation (ICE) curves. As of this writing, the concept is experimental. Caveat emptor!
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
R
man
.Rbuildignore
.gitignore
DESCRIPTION
LICENSE
NAMESPACE
README.md
maRginal.Rproj

README.md

marginal will be an R package for calculating marginal effects for arbitrary prediction models. As of now, code and research are in early stages.

Open issues include:

  • Finding a valid way to calculate confidence intervals
  • Handling categorical variables more elegantly
  • Finding the best default behaviors for calculations
  • And more!

Some basic usage

devtools::install_github("tommyjones/marginal")

library(marginal)
library(randomForest)

data(mtcars)

fit <- randomForest(mpg ~., data = mtcars)

mfx <- CalcMfx(object = fit, X = mtcars)

plot(mfx)