Predictions from an ares fit — predict.ares • roadrunner

Returns predictions for new data from a fitted ares MARS model. For bagged fits (n.boot > 0), returns the mean across the central fit and the bootstrap replicates; with se.fit = TRUE, the per-row bag standard deviation is attached as an "sd" attribute. For gaussian fits built with a residual variance model (varmod = "const" or "lm"), interval = "pint" returns prediction intervals.

Usage

# S3 method for class 'ares'
predict(
  object,
  newdata = NULL,
  type = c("response", "link"),
  se.fit = FALSE,
  interval = c("none", "pint"),
  level = 0.95,
  ...
)

Arguments

object

An object of class "ares".

newdata

A numeric matrix or data frame of new predictors. NULL (default) returns object$fitted.values.

type

Prediction scale.

"response" (default): probabilities for binomial, response-scale means for poisson and gamma, fitted values for gaussian.
"link": the linear predictor. Coincides with "response" for gaussian fits.

se.fit

If TRUE and the fit was built with n.boot > 0, the returned vector carries an "sd" attribute holding the per-row bag standard deviation. Default FALSE.

interval

Type of interval to return.

"none" (default): return point predictions only.
"pint": return a matrix with columns c("fit", "lwr", "upr") using the variance model stored at fit time. Requires family = "gaussian" and varmod = "const" or "lm"; errors otherwise.

level

Confidence level for prediction intervals when interval = "pint". Default 0.95.

...

Currently unused.

Value

A numeric vector of predictions, or a matrix with columns c("fit", "lwr", "upr") when interval = "pint". With se.fit = TRUE on a bagged fit, the vector carries an "sd" attribute.

Examples

fit <- ares(as.matrix(mtcars[, -1]), mtcars$mpg, nthreads = 2)
p <- predict(fit, as.matrix(mtcars[, -1]))