SuperLearner: predict.superlearner – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

predict.superlearner

Predict method for SuperLearner object

Description

Obtains predictions on a new data set from a SuperLearner fit. May require the original data if one of the library algorithms uses the original data in its predict method.

Usage

## S3 method for class 'SuperLearner'
predict(object, newdata, X = NULL, Y = NULL,
  onlySL = FALSE, ...)

Arguments

`object`	Fitted object from `SuperLearner`
`newdata`	New X values for prediction
`X`	Original data set used to fit `object`, if needed by fit object.
`Y`	Original outcome used to fit `object`, if needed by fit object.
`onlySL`	Logical. If TRUE, only compute predictions for algorithms with non-zero coefficients in the super learner object. Default is FALSE (computes predictions for all algorithms in library).
`...`	Additional arguments passed to the `predict.SL.*` functions

Details

If newdata is omitted the predicted values from object are returned. Each algorithm in the Super Learner library needs to have a corresponding prediction function with the “predict.” prefixed onto the algorithm name (e.g. predict.SL.glm for SL.glm).

Value

`pred`	Predicted values from Super Learner fit
`library.predict`	Predicted values for each algorithm in library

Author(s)

Eric C Polley epolley@uchicago.edu