create.restricted.models: Creates restricted models for IKDE

In tkmckenzie/ikde: Iterative Kernel Density Estimation

Description

Creates set of restricted models to be used for posterior density estimation

Usage

create.restricted.models(ikde.model, eval.point)

Arguments

ikde.model	An object of class ikde.model, does not necessarily have to be built
eval.point	A list of parameter names and the point to evaluate densities

Details

Posterior density can be estimated by breaking the multi-dimensional density into one-dimensional components. This method creates restricted models from which conditional densities can be estimated. Each real parameter and each entry of vector parameters are restricted one at a time, with values restricted at the specified point.

Value

Returns a list of built ikde.models for each restricted model

Examples

data(lm.generated)

X <- lm.generated$X
y <- lm.generated$y

data <- list(N = list(type = "int<lower=1>", dim = 1, value = nrow(X)),
             k = list(type = "int<lower=1>", dim = 1, value = ncol(X)),
             X = list(type = "matrix", dim = "[N, k]", value = X),
             y = list(type = "vector", dim = "[N]", value = y))
parameters <- list(beta = list(type = "vector", dim = "[k]"),
                   sigma_sq = list(type = "real<lower=0>", dim = 1))
model <- list(priors = c("beta ~ normal(0, 10);",
                         "sigma_sq ~ inv_gamma(1, 1);"),
              likelihood = c("y ~ normal(X * beta, sqrt(sigma_sq));"))

ikde.model <- define.model(data, parameters, model)
eval.point <- list(beta = c(1, 2, 3, 4),
                   sigma = 5)

ikde.model.list <- create.restricted.models(ikde.model, eval.point)
for (restricted.ikde.model in ikde.model.list){
  cat(restricted.ikde.model$stan.code)
  cat("--------------------------------------------------\n")
}

tkmckenzie/ikde documentation built on May 13, 2019, 9:53 p.m.