matern_pc_prior: Helper funcion to specify a Penalized Complexity (PC) prior... In SpatialGEV: Fit Spatial Generalized Extreme Value Models

 matern_pc_prior R Documentation

Helper funcion to specify a Penalized Complexity (PC) prior on the Matern hyperparameters

Description

Helper funcion to specify a Penalized Complexity (PC) prior on the Matern hyperparameters

Usage

``````matern_pc_prior(rho_0, p_rho, sig_0, p_sig)
``````

Arguments

 `rho_0` Hyperparameter for PC prior on the range parameter. Must be positive. See details. `p_rho` Hyperparameter for PC prior on the range parameter. Must be between 0 and 1. See details. `sig_0` Hyperparameter for PC prior on the scale parameter. Must be positive. See details. `p_sig` Hyperparameter for PC prior on the scale parameter. Must be between 0 and 1. See details.

Details

The joint prior on `rho` and `sig` achieves

```P(rho < rho_0) = p_rho,
```

and

```P(sig > sig_0) = p_sig,
```

where `rho = sqrt(8*nu)/kappa`.

Value

A list to provide to the `matern_pc_prior` argument of `spatialGEV_fit`.

References

Simpson, D., Rue, H., Riebler, A., Martins, T. G., & SĂ¸rbye, S. H. (2017). Penalising model component complexity: A principled, practical approach to construct priors. Statistical Science.

Examples

``````
n_loc <- 20
y <- simulatedData2\$y[1:n_loc]
locs <- simulatedData2\$locs[1:n_loc,]
fit <- spatialGEV_fit(
data = y,
locs = locs,
random = "abs",
init_param = list(
a = rep(0, n_loc),
log_b = rep(0, n_loc),
s = rep(-2, n_loc),
beta_a = 0,
beta_b = 0,
beta_s = -2,
log_sigma_a = 0,
log_kappa_a = 0,
log_sigma_b = 0,
log_kappa_b = 0,
log_sigma_s = 0,
log_kappa_s = 0
),
reparam_s = "positive",
kernel = "matern",
beta_prior = list(
beta_a=c(0,100),
beta_b=c(0,10),
beta_s=c(0,10)
),
matern_pc_prior = list(
matern_a=matern_pc_prior(1e5,0.95,5,0.1),
matern_b=matern_pc_prior(1e5,0.95,3,0.1),
matern_s=matern_pc_prior(1e2,0.95,1,0.1)
)
)

``````

SpatialGEV documentation built on June 22, 2024, 9:24 a.m.