decodeLDA {rDecode}

Implement DECODE for simple LDA

Description

Implement DECODE for simple LDA. The LDA assumes both classes have equal prior probabilities. This implementation is used in Hadimaja and Pun (2018).

Usage

decodeLDA(X, y, lambda0 = NULL, ...)

Arguments

X

n\times p data matrix.

y

binary n-length vector containing the class of each observation.

lambda0

number between 0 and 1. If NULL, will use \sqrt{2 \log{p}/n}.

...

additional arguments to be passed to general decode function.

Value

An object of class decodeLDA containing:

eta

DECODE of \Omega\delta

X

training data used

y

training label used

and various outputs from decode function.

References

Hadimaja, M. Z., & Pun, C. S. (2018). A Self-Calibrated Regularized Direct Estimation for Graphical Selection and Discriminant Analysis.

Examples

# for efficiency, we will only use 500 variables

# load the training data (Lung cancer data, cleaned)
data(lung.train) # 145 x 1578
X.train <- lung.train[,1:500]
y.train <- lung.train[,1578]

# build the DECODE
object <- decodeLDA(X.train, y.train)

object
summary(object)
coef(object)

# test on test data
data(lung.test)
X.test <- lung.test[,1:500]
y.test <- lung.test[,1578]
y.pred <- predict(object, X.test)
table(y.pred, y.test)