transform: Transform hyperparameters in multivarite spatial models.

In INLAMSM: Multivariate Spatial Models with 'INLA'

Description

Multivariate spatial models fit will report hyperparameters in the internal scale. These functions will transform the hyperparameters to a different scale. Using this function requires setting control.compute = list(config = TRUE) when fitting the model with INLA.

Usage

inla.MCAR.transform(obj, k, model = "IMCAR", alpha.min, alpha.max)

Arguments

obj	An 'inla' object with an MCAR, IMCAR or M-model latent effect.
k	Number of variables in the multivariate model.
model	Either "INDIMCAR", "INDPMCAR", "IMCAR" or "PMCAR". Not used for M-models.
alpha.min	Lower bound of the autocorrelation parameter alpha.
alpha.max	Upper bound of the autocorrelation parameter alpha.

Value

This function returns a list with the following elements:

• marginals.hyperpar List with the posterior marginals of transformed hyperparameters.

• summary.hyperpar Summary of the posterior marginals.

• VAR.p Variance matrix of between-variables variability computed using point estimates from the posterior marginals.

• VAR.m Posterior mean of variance matrix of the between-variables variability. This is computed using the internal representation of the posterior joint distribution of the hyperparameters.

• confs Configurations of the hyperparameters used to compute VAR.m. This is obtained from obj$misc$configs$config.

• M.p M matrix (only in the M-model) obtained using point estimates of the parameters.

• M.m M matrix (only in the M-model) obtained using the the internal representation of the posterior joint distribution of the hyperparameters.

References

Palmí-Perales F, Gómez-Rubio V, Martinez-Beneito MA (2021). “Bayesian Multivariate Spatial Models for Lattice Data with INLA.” _Journal of Statistical Software_, *98*(2), 1-29. doi: 10.18637/jss.v098.i02 (URL: https://doi.org/10.18637/jss.v098.i02).

INLAMSM documentation built on June 4, 2021, 9:07 a.m.