Skip to contents

Posterior predictive distribution from a Pillar 2 deep ensemble

Source: R/piml_posterior.R
piml_neural_ode_posterior.Rd

Convenience wrapper that calls predict.edaphos_piml_neural_ode_ensemble() on the requested depths and returns the result as an edaphos_posterior() with query_type = "sample" (one scalar prediction per query depth) or "feature" (when the caller passes a single depth and wants the posterior over that single scalar).

Usage

piml_neural_ode_posterior(
  object,
  newdepths,
  include_obs_noise = FALSE,
  seed = NULL,
  units = NULL
)

Arguments

object

An edaphos_piml_neural_ode_ensemble.

newdepths

Numeric vector of depths at which to evaluate the posterior predictive.

include_obs_noise

Logical — when TRUE, adds Gaussian observation noise to every draw so the posterior is the predictive distribution of a future observation (aleatoric included) rather than only the uncertainty on the mean function. Default FALSE.

seed

Optional RNG seed for the observation-noise draw.

units

Optional units tag.

Value

An edaphos_posterior with method = "ensemble".