fit.model: Fits Stan model

In tkmckenzie/ikde: Iterative Kernel Density Estimation

Description

Uses a built ikde.model to draw samples from posterior distribution using Stan

Usage

fit.model(ikde.model, burn.iter = 1000, sample.iter = 1000,
  chains = 1, control = NULL, refresh = NULL,
  display.output = FALSE)

Arguments

ikde.model	An object of class ikde.model which has been built
burn.iter	Number of warmup iterations
sample.iter	Number of sampling iterations
chains	Number of independent chains to use
control	Control parameters used in the Markov chain. See ?rstan::stan for details.
refresh	How frequently should progress be reported, in numbers of iterations
display.output	Boolean indicating whether output from rstan::stan should be printed

Details

Takes a built ikde.model object, which contains model DSO, and fits the model using rstan::stan.

Value

An object of S4 class stanfit. See rstan::stan for more details.

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)
ikde.model <- build.model(ikde.model)
stan.fit <- fit.model(ikde.model)
stan.extract <- extract(stan.fit)

# Only an estimation, may not exactly match presented result
print(apply(stan.extract$beta, 2, mean))
# [1] 3.199021 1.620546 4.489716 1.226508

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