cv.stepSplitReg: Cross Validation - Stepwise Split Regularized Regression

In stepSplitReg: Stepwise Split Regularized Regression

Cross Validation - Stepwise Split Regularized Regression

Description

cv.stepSplitReg performs the CV procedure for stepwise split regularized regression.

Usage

cv.stepSplitReg(
  x,
  y,
  n_models = NULL,
  max_variables = NULL,
  keep = 1,
  model_criterion = c("F-test", "RSS")[1],
  stop_criterion = c("F-test", "pR2", "aR2", "R2", "Fixed")[1],
  stop_parameter = 0.05,
  shrinkage = TRUE,
  alpha = 3/4,
  include_intercept = TRUE,
  n_lambda = 100,
  tolerance = 0.001,
  max_iter = 1e+05,
  n_folds = 10,
  model_weights = c("Equal", "Proportional", "Stacking")[1],
  n_threads = 1
)

Arguments

x	Design matrix.
y	Response vector.
n_models	Number of models into which the variables are split.
max_variables	Maximum number of variables that a model can contain.
keep	Proportion of models to keep based on their individual cross-validated errors. Default is 1.
model_criterion	Criterion for adding a variable to a model. Must be one of c("F-test", "RSS"). Default is "F-test".
stop_criterion	Criterion for determining when a model is saturated. Must be one of c("F-test", "pR2", "aR2", "R2", "Fixed"). Default is "F-test".
stop_parameter	Parameter value for the stopping criterion. Default is 0.05 for "F-test".
shrinkage	TRUE or FALSE parameter for shrinkage of the final models. Default is TRUE.
alpha	Elastic net mixing parmeter for model shrinkage. Default is 3/4.
include_intercept	TRUE or FALSE parameter for the inclusion of an intercept term.
n_lambda	Number of candidates for the sparsity penalty parameter. Default is 100.
tolerance	Convergence criteria for the coefficients. Default is 1e-3.
max_iter	Maximum number of iterations in the algorithm. Default is 1e5.
n_folds	Number of cross-validation folds. Default is 10.
model_weights	Criterion to determine the weights of the model for prediciton. Must be one of c("Equal", "Proportional", "Stacking"). Default is "Equal".
n_threads	Number of threads. Default is 1.

Value

An object of class cv.stepSplitReg.

Author(s)

Anthony-Alexander Christidis, [e-mail]

See Also

coef.cv.stepSplitReg, predict.cv.stepSplitReg

Examples

# Required Libraries
library(mvnfast)

# Setting the parameters
p <- 100
n <- 30
n.test <- 500
sparsity <- 0.2
rho <- 0.5
SNR <- 3

# Generating the coefficient
p.active <- floor(p*sparsity)
a <- 4*log(n)/sqrt(n)
neg.prob <- 0.2
nonzero.betas <- (-1)^(rbinom(p.active, 1, neg.prob))*(a + abs(rnorm(p.active)))

# Correlation structure
Sigma <- matrix(0, p, p)
Sigma[1:p.active, 1:p.active] <- rho
diag(Sigma) <- 1
true.beta <- c(nonzero.betas, rep(0 , p - p.active))

# Computing the noise parameter for target SNR
sigma.epsilon <- as.numeric(sqrt((t(true.beta) %*% Sigma %*% true.beta)/SNR))

# Simulate some data
set.seed(1)
x.train <- mvnfast::rmvn(n, mu=rep(0,p), sigma=Sigma)
y.train <- 1 + x.train %*% true.beta + rnorm(n=n, mean=0, sd=sigma.epsilon)
x.test <- mvnfast::rmvn(n.test, mu=rep(0,p), sigma=Sigma)
y.test <- 1 + x.test %*% true.beta + rnorm(n.test, sd=sigma.epsilon)

# Stepwise Split Regularized Regression
step.out <- cv.stepSplitReg(x.train, y.train, n_models = c(2, 3), max_variables = NULL, keep = 4/4,
                            model_criterion = c("F-test", "RSS")[1],
                            stop_criterion = c("F-test", "pR2", "aR2", "R2", "Fixed")[1], 
                            stop_parameter = 0.05, 
                            shrinkage = TRUE, alpha = 4/4, include_intercept = TRUE, 
                            n_lambda = 50, tolerance = 1e-2, max_iter = 1e5, n_folds = 5, 
                            model_weights = c("Equal", "Proportional", "Stacking")[1], 
                            n_threads = 1)
step.coefficients <- coef(step.out, group_index = 1:step.out$n_models_optimal)
step.predictions <- predict(step.out, x.test, group_index = 1:step.out$n_models_optimal)
mspe.step <- mean((step.predictions-y.test)^2)/sigma.epsilon^2

stepSplitReg documentation built on Nov. 22, 2022, 5:08 p.m.