Predict Causal Effects — predict.frengression

Computes predictions from the causal margin P(Y|do(X=x)), returning the mean, quantiles, or raw samples. For the learned model, this marginalizes over the noise input eta ~ N(0,I) that substitutes for the confounders.

Usage

# S3 method for class 'frengression_seq'
predict(object, s, x, type = "mean", nsample = 200, ...)

# S3 method for class 'frengression_surv'
predict(object, s, x, type = "mean", nsample = 200, ...)

# S3 method for class 'frengression'
predict(
  object,
  x,
  type = c("mean", "sample", "quantile"),
  nsample = 200,
  quantiles = seq(0.1, 0.9, 0.1),
  trim = 0.05,
  ...
)

Arguments

object: A frengression object.
s: Static baseline variables (matrix or tensor).
x: Intervention values for X (matrix, vector, or data frame).
type: Type of prediction: "mean", "quantile", or "sample" (default: "mean").
nsample: Number of Monte Carlo samples (default: 200).
...: Additional arguments (ignored).
quantiles: Quantile levels if type = "quantile" (default: seq(0.1, 0.9, 0.1)).
trim: Trimming proportion for mean (default: 0.05).

Value

A matrix of predictions.