evaluate.likelihood: Stan model likelihood evaluation

In tkmckenzie/ikde: Iterative Kernel Density Estimation

Description

Evaluates likelihood of Stan model at specified evaluation point

Usage

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

Arguments

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 likelihood

Details

Parses sampling statements in ikde.model$model$likelihood and evaluates them at the specified evaluation point.

Value

A real number indicating value of the log-likelihood at the specified 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)

eval.point <- list(beta = c(1, 2, 3, 4), sigma_sq = 5)

# These results match:
evaluate.likelihood(ikde.model, eval.point)
sum(dnorm(y, X %*% eval.point$beta, eval.point$sigma_sq, log = TRUE))
# [1] -1054.093
 

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