Kriging prediction for a hybrid Whittle-Matern SPDE model

Computes posterior predictions for the hybrid model \(Y_i = (A Y)_i + \epsilon_i\), where \(Y\) satisfies \(L^{\alpha/2}(\tau Y) = \beta X + W\). The posterior of the centred latent field \(Y - \mu\) is computed using the underlying SPDE prediction routine; the deterministic mean \(\mu\) is then added back at the prediction locations.

Usage

# S3 method for class 'hybrid_spde'
predict(
  object,
  A,
  Aprd,
  Y,
  sigma.e,
  compute.variances = FALSE,
  posterior_samples = FALSE,
  n_samples = 100,
  only_latent = FALSE,
  ...
)

Arguments

object

A hybrid_spde object.

A

Projection matrix linking the observations to the mesh nodes.

Aprd

Projection matrix linking the prediction locations to the mesh nodes.

Y

Observations.

sigma.e

Standard deviation of the measurement noise.

compute.variances

If TRUE, the kriging variances are returned.

posterior_samples

If TRUE, posterior samples are returned.

n_samples

Number of posterior samples.

only_latent

If TRUE, the posterior samples are not perturbed by the measurement noise.

...

Additional arguments passed to the underlying predict method.

Value

A list with components mean, optionally variance and samples.

See also

hybrid.spde()