Skip to contents

Posterior predictive distribution of a Bayesian Pillar 2 fit

Source: R/piml_bayesian.R
predict.edaphos_piml_bayes.Rd

Propagates the full posterior over ODE parameters through the forward model to produce posterior-predictive draws of \(y(z)\) at user-supplied depths.

Usage

# S3 method for class 'edaphos_piml_bayes'
predict(
  object,
  newdepths,
  n_draws = NULL,
  interval = NULL,
  include_obs_noise = FALSE,
  seed = NULL,
  ...
)

Arguments

object

An edaphos_piml_bayes returned by piml_profile_fit_bayesian().

newdepths

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

n_draws

Integer — number of posterior draws to propagate. Defaults to min(500, nrow(object$draws)).

interval

Optional numeric scalar in (0, 1). When given, returns a tidy data.frame with mean, sd, lower, upper per depth (symmetric central credible interval). When NULL (default), returns the full n_draws-by-length(newdepths) matrix of predictive draws.

include_obs_noise

Logical — if TRUE, Gaussian observation noise N(0, sigma^2) is added to every predictive draw so the interval represents the predictive distribution of a future observation. If FALSE (default), the interval represents the uncertainty on the mean function \(g(z; \theta)\) alone.

seed

Optional integer seed for the sub-sampling of draws.

...

Unused; present for S3 predict generic compatibility.

Value

Either a matrix (when interval is NULL) or a data frame with columns depth, mean, sd, lower, upper.

Examples

depths <- c(5, 15, 30, 60, 100)
values <- c(25, 18, 12, 8, 6.5)
fit <- piml_profile_fit_bayesian(depths, values)
predict(fit, newdepths = c(10, 20, 40, 80), interval = 0.95)
#>   depth      mean        sd     lower     upper
#> 1    10 21.012520 0.1387745 20.742415 21.278981
#> 2    20 15.504483 0.1367901 15.214314 15.755396
#> 3    40 10.069634 0.1481595  9.800381 10.377096
#> 4    80  7.029437 0.1405769  6.787625  7.327181