Examples · CmdStanPHP

The package ships three example models from CmdStanR, with their data: logistic, a logistic regression, and schools and schools_ncp, the eight schools hierarchical model in its centered and non-centered forms. The repository’s examples/ directory has the small programs used in these docs.

Bundled models

cmdstanr_example() compiles an example, fits it with its bundled data, and returns the fit. The first argument names the example, and the second the method: sample (the default), optimize, laplace, variational, pathfinder, or diagnose. Other arguments are passed to the method. The executable is built in the directory set by the cmdstanphp_example_dir option, or in cmdstanphp-examples under the system temporary directory. print_example_program() prints the Stan program.

CmdStan\print_example_program('schools_ncp');

$fit = CmdStan\cmdstanr_example('schools_ncp', seed: 123, show_messages: false);
print_r($fit->summary(['mu', 'tau'], measures: ['mean', 'sd', 'rhat', 'ess_bulk']));
print_r($fit->diagnosticWarnings());

data {
  int<lower=1> J;
  vector<lower=0>[J] sigma;
  vector[J] y;
}
parameters {
  real mu;
  real<lower=0> tau;
  vector[J] theta_raw;
}
transformed parameters {
  vector[J] theta = mu + tau * theta_raw;
}
model {
  target += normal_lpdf(tau | 0, 10);
  target += normal_lpdf(mu | 0, 10);
  target += normal_lpdf(theta_raw | 0, 1);
  target += normal_lpdf(y | theta, sigma);
}
Array
(
    [0] => Array
        (
            [variable] => mu
            [mean] => 6.6143359951631
            [sd] => 4.2648286711691
            [rhat] => 1.0014001162774
            [ess_bulk] => 3501.9077743615
        )

    [1] => Array
        (
            [variable] => tau
            [mean] => 4.710142117602
            [sd] => 3.6457106662655
            [rhat] => 1.0004923423035
            [ess_bulk] => 1748.8566170691
        )

)
Array
(
)

The centered form, schools, samples the same posterior with theta as a parameter. With the same seed it produces divergent transitions in every chain:

$fit = CmdStan\cmdstanr_example('schools', seed: 123, show_messages: false);
print_r($fit->diagnosticWarnings());

Array
(
    [0] => Chain 1: 73 divergent transitions.
    [1] => Chain 2: 74 divergent transitions.
    [2] => Chain 2: E-BFMI below 0.3.
    [3] => Chain 3: 25 divergent transitions.
    [4] => Chain 4: 8 divergent transitions.
)

The MCMC guide explains these warnings. The outputs above are from CmdStan 2.40.0 on macOS. Other platforms and CmdStan versions give different draws for the same seed.

Programs in the repository

The examples/ directory of the repository holds:

bernoulli.stan and bernoulli.php, the model from Getting started and a script that samples it. Run php examples/bernoulli.php from a clone of the repository after composer install.
generated.stan, the Bernoulli model with a posterior prediction, used in Generated quantities.
normal.stan, a one-parameter normal model, used in Log density and model methods.