predict.asgl: Make predictions from an 'asgl' object

In jeffdaniel/asgl: Fit a GLM with Adaptive Sparse Group Lasso Penalty

Description

Makes predictions from a fitted 'asgl' object.

Usage

## S3 method for class 'asgl'
predict(object, newx, s = NULL, type = c("link",
  "response"), ...)

Arguments

object	Fitted asgl object
newx	Matrix of new values for x at which predictions are to be made.
s	Value(s) of the penalty tuning parameter at which predictions are to be made. Default is the entire sequence used to fit the regularization path.
type	Type of predicted required. For "gaussian" models, "link" and "response" give the same result: the predicted response. For "binomial" models, "link" gives the linear predictor and "response" gives the predicted probabilities.
...	Ignored.

Value

A matrix of predictions or, if s is of length 1, a vector of predictions.

Examples

# linear regression
n <- 500; p <- 20; groupsize <- 5
index <- ceiling(1:p / groupsize)
beta <- (-2:2)
x <- matrix(rnorm(n * p), ncol = p, nrow = n)
y <- as.vector(x[,1:5] %*% beta + 0.1 * rnorm(n))
fit1 <- asgl(x, y, index, family = "gaussian")
predict(fit1, x)

# logistic regression
eta <- x[, 1:5] %*% beta
prob <- exp(eta) / (1 + exp(eta))
y <- rbinom(n, 1, prob)
fit2 <- asgl(x, y, index, family = "binomial")
predict(fit2, x, s = fit2$lambda[20], type = "response")

jeffdaniel/asgl documentation built on Nov. 4, 2019, 2:37 p.m.