Cauchy
HOBBS form: x(i) ~ dcauchy(location, scale);
The package test file performs:
- dcauchy recovers its location
- dcauchy gives the correct posterior for its location
- dcauchy matches Stan and JAGS
Same model in HOBBS, Stan, and JAGS
param loc(1);
block loc(1) {
loc(1) ~ dnorm(0, 3);
for (i = 1:n) y(i) ~ dcauchy(loc(1), 1.0);
}
data {
int<lower=1> n;
vector[n] y;
}
parameters {
vector[1] loc;
}
model {
loc[1] ~ normal(0, 3);
y ~ cauchy(loc[1], 1);
}model {
loc[1] ~ dnorm(0, 0.1111111111111111)
for (i in 1:n) { y[i] ~ dt(loc[1], 1, 1) }
}
Run this test
From the documentation project itself:
library(hobbs)
library(testthat)
source("validation/source/tests/testthat/helper-distributions.R")
testthat::test_file("validation/source/tests/testthat/test-dist-cauchy.R")From the HOBBS package source tree, the normal package test workflow is simpler:
devtools::test(filter = "dist-cauchy")Complete test file
test_that("dcauchy recovers its location", {
set.seed(109); truth <- 0.65; y <- rcauchy(90, truth, 1)
model <- 'param loc(1);
block loc(1) {
loc(1) ~ dnorm(0, 3);
for (i = 1:n) y(i) ~ dcauchy(loc(1), 1.0);
}'
d <- hobbs_test_draws(model, list(y = y)); expect_posterior_near(d, "loc[1]", truth, 0.38)
})
test_that("dcauchy gives the correct posterior for its location", {
set.seed(109)
truth <- 0.65
y <- rcauchy(90, truth, 1)
model <- 'param loc(1);
block loc(1) {
loc(1) ~ dnorm(0, 3);
for (i = 1:n) y(i) ~ dcauchy(loc(1), 1.0);
}'
d <- hobbs_test_draws(model, list(y = y))
log_posterior <- function(loc) {
dnorm(loc, 0, 3, log = TRUE) +
sum(dcauchy(y, loc, 1, log = TRUE))
}
expect_numerical_posterior(
d, "loc[1]", log_posterior,
lower = -2.5, upper = 2.5,
mean_tolerance = 0.05, sd_tolerance = 0.04
)
})
test_that("dcauchy matches Stan and JAGS", {
skip_if_reference_samplers_missing()
set.seed(109)
truth <- 0.65
y <- rcauchy(90, truth, 1)
n <- length(y)
hobbs_model <- 'param loc(1);
block loc(1) {
loc(1) ~ dnorm(0, 3);
for (i = 1:n) y(i) ~ dcauchy(loc(1), 1.0);
}'
stan_model <- '
data {
int<lower=1> n;
vector[n] y;
}
parameters {
vector[1] loc;
}
model {
loc[1] ~ normal(0, 3);
y ~ cauchy(loc[1], 1);
}'
jags_model <- '
model {
loc[1] ~ dnorm(0, 0.1111111111111111)
for (i in 1:n) { y[i] ~ dt(loc[1], 1, 1) }
}'
data <- list(n = n, y = y)
d_hobbs <- hobbs_test_draws(hobbs_model, list(y = y))
d_stan <- stan_test_draws(stan_model, data, "loc")
d_jags <- jags_test_draws(jags_model, data, "loc")
expect_posterior_matches_reference(d_hobbs, d_stan, d_jags, "loc[1]")
})