Predict from the sparse or dual SDR engine — predict.risdr

Projects new observations using the training centring, scaling, and selected directions. Survival predictions are Cox linear predictors or relative risks; they are not predicted survival times.

Usage

# S3 method for class 'risdr_sparse'
predict(
  object,
  newX,
  type = c("response", "scores", "link", "risk", "class", "probs"),
  ...
)

# S3 method for class 'risdr_dual'
predict(
  object,
  newX,
  type = c("response", "scores", "link", "risk", "class", "probs"),
  ...
)

Arguments

object: A sparse or dual RISDR fit.
newX: Numeric predictor matrix or data frame, with named columns.
type: Prediction type. "response" returns continuous predictions, binary probabilities, multiclass labels, or survival relative risks. "scores" returns the selected SDR coordinates. "link" and "risk" are available for survival outcomes. "class" and "probs" are available for categorical outcomes.
...: Additional arguments passed to the downstream prediction method.

Value

Predictions or a score matrix, according to type.