Elicit a prior for GGM parameters

Elicit a prior for GGM parameters

Usage

# S3 method for class 'ggm_parameters'
elicit_prior(params, prior = NULL, ...)

Arguments

params

A ggm_parameters object, as returned by ggm_parameters. It also inherits from bgm_parameters.

prior

Character scalar naming the precision prior, or NULL (default) to infer it from the graph: "wishart" for a complete G, "gwishart" for a sparse G. Supplying prior overrides the default and is validated against the graph (e.g. a sparse-only prior cannot be placed on a complete graph).

...

Family-specific arguments.

Value

A bgm_elicited object describing the elicited prior. It also carries a family-specific class (for example ggm_elicited), on which the downstream methods dispatch.

Details

The elicitation currently applies the K/nu centering, which is exact for Wishart, and approximate for G-Wishart.

See also

elicit_prior() for the generic method and prior_ess() for prior ESS estimation.

Other prior elicitation: elicit_prior(), ggm_study(), simulate_prior_study(), simulate_prior_study.ggm_study()

Examples

p <- 3
G <- matrix(0, p, p)
G[1, 2] <- G[2, 1] <- 1
K <- diag(p)
K[1, 2] <- K[2, 1] <- 0.3
params <- ggm_parameters(K = K, G = G, nu = 10)
ep <- elicit_prior(params)
ep
#> <ggm_elicited>  prior: gwishart 
#>   nodes : 3 
#>   edges : 1 of 3 
#>   nu    : 10  (prior study size)