The R package DAP
provides tools for high-dimensional
binary classification in the case of unequal covariance matrices. It
implements methods from the following paper: * Sparse quadratic classification
rules via linear dimension reduction by Gaynanova and Wang
(2017).
To install the latest version from Github, use
devtools::install_github("irinagain/DAP")
(DAP)
library(MASS)
library
# Example
## Specify model parameters
= 100
p = rep(0, p)
mu1 = c(rep(3, 10), rep(0, p-10))
mu2 = diag(p)
Sigma1 = 0.5*diag(p)
Sigma2
## Build training data and test data
= 50
n_train = 50
n_test = MASS::mvrnorm(n = n_train, mu = mu1, Sigma = Sigma1)
x1 = MASS::mvrnorm(n = n_train, mu = mu2, Sigma = Sigma2)
x2 = rbind(x1, x2)
xtrain = MASS::mvrnorm(n = n_test, mu = mu1, Sigma = Sigma1)
x1_test = MASS::mvrnorm(n = n_test, mu = mu2, Sigma = Sigma2)
x2_test xtest = rbind(x1_test, x2_test)
= c(rep(1, n_train), rep(2, n_train))
ytrain = c(rep(1, n_test), rep(2, n_test))
ytest
## Apply DAP
# Given ytest, the function returns the miclassification error rate.
= apply_DAP(xtrain, ytrain, xtest, ytest)
ClassificationError
# Without ytest, the function returns predicted labels.
= apply_DAP(xtrain, ytrain, xtest) Ypredict
This package is free and open source software, licensed under GPL (>=2).