Description of R package MGSDA

MGSDA (Multi-Group Sparse Discriminant Analysis) is an R package that implements methods described in

The package is available from CRAN.

To install from Github:


The main functions are cv.dLDA(cross-validation), dLDA(fitting for specified value of tuning parameter) and classifyV(classification). Each function has a documentation with a simple example which can be accessed using standard ? commands in R (i.e. ?cv.dLDA).

Please feel free to contact me at irinag [at] stat [dot] tamu [dot] edu if you have any questions or experience problems with the package.

Simple example


### Example 1
# generate training data
n <- 10
p <- 100
G <- 3
ytrain <- rep(1:G, each = n)
xtrain <- matrix(rnorm(p * n * G), n * G, p)

# find matrix of canonical vectors V
V <- dLDA(xtrain, ytrain, lambda = 0.1)
sum(rowSums(V) != 0)

# generate test data
m <- 20
xtest <- matrix(rnorm(p * m), m, p)

# perform classification
ytest <- classifyV(xtrain, ytrain, xtest, V)