predict.cv.missoNet {missoNet}

Make predictions from a cv.missoNet object

Description

S3 method for making predictions of response values from a fitted 'cv.missoNet' object.

Usage

## S3 method for class 'cv.missoNet'
predict(object, newx = NULL, s = "lambda.min", ...)

Arguments

object

A fitted 'cv.missoNet' object.

newx

A predictor matrix of new values at which predictions are to be made. The columns of 'newx' should have the same standardization flags as the original input for training the model. Missing values are not allowed. 'newx' should not include a column of ones for an intercept.

s

Character string, the regularization parameter pair \lambda = (\lambda_B, \lambda_\Theta) at which the coefficients are extracted for making predictions. It supports three special strings, named "lambda.min" (default), "lambda.1se.Beta" and "lambda.1se.Theta".

...

Not used. Other arguments for predicting.

Value

The matrix of predicted values: 'newy = mu_hat + newx %*% Beta_hat'.

Author(s)

Yixiao Zeng [e-mail], Celia M.T. Greenwood and Archer Yi Yang.

Examples

## Simulate a dataset.
set.seed(123)  # reproducibility
sim.dat <- generateData(n = 300, p = 10, q = 10, rho = 0.1, missing.type = "MCAR")
tr <- 1:240  # training set indices
tst <- 241:300  # test set indices


## Perform a five-fold cross-validation on the training set.
cvfit <- cv.missoNet(X = sim.dat$X[tr, ], Y = sim.dat$Z[tr, ], kfold = 5,
                     fit.1se = TRUE, permute = TRUE, with.seed = 486)


## Make predictions of response values on the test set.
newy1 <- predict(cvfit, newx = sim.dat$X[tst, ], s = "lambda.min")
newy2 <- predict(cvfit, newx = sim.dat$X[tst, ], s = "lambda.1se.Beta")  # 'fit.1se' = TRUE
newy3 <- predict(cvfit, newx = sim.dat$X[tst, ], s = "lambda.1se.Theta")  # 'fit.1se' = TRUE