evaluate.statement: Evaluate sampling statement from Stan program

In tkmckenzie/ikde: Iterative Kernel Density Estimation

Description

Evaluate sampling statement from Stan program

Usage

evaluate.statement(statement, ikde.model, eval.point)

Arguments

statement	A string containing a sampling statement
ikde.model	An object of class ikde.model, which has been built
eval.point	A list of parameter names and the point to evaluate the statement

Details

Parses the given sampling statement and evaluates it at the specified evaluation point. The ikde.model object and eval.point object are needed to resolve variable values in the statement.

Value

A real number indicating value of the log-density of the statement at the evaluation point

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)

statement <- ikde.model$model$likelihood[1]
eval.point <- list(beta = c(1, 2, 3, 4), sigma_sq = 5)

# These results match:
evaluate.statement(statement, ikde.model, eval.point)
sum(dnorm(y, mean = X %*% eval.point$beta, sd = sqrt(eval.point$sigma_sq), log = TRUE))
# [1] -4178.641
  

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