phpstats / cmdstanphp

v0.1.2 2026-10-31 15:26 UTC

Requires

• php: >=8.2 <8.6
• ext-json: *

Requires (Dev)

• phpunit/phpunit: ^11.5
• squizlabs/php_codesniffer: ^3.13

Suggests

• ext-ffi: Required for model methods (log_prob, gradients, constrain/unconstrain) and expose_functions()
• ext-phit: Required for loo(moment_match: true): provides the Phit\draws_moments, Phit\draws_covariance_map, and Phit\draws_affine kernels; StanData::fromFormula() also needs its Phit\model_matrix()
• phpstats/phibble: Data frames: pass a Phibble\DataFrame as model data, and get draws and summaries back with drawsTable() and summaryTable() (phibble requires PHP 8.3 or later)

MIT

Published on Packagist.org on 2026-10-31 17:02 UTC

README

Compile Stan models, run Bayesian inference, and inspect posterior draws from PHP.

CmdStanPHP runs Stan programs through CmdStan, Stan's command-line interface. It is a port of CmdStanR 1.0 and keeps CmdStanR's function and option names.

Documentation: https://cmdstanphp.phpstats.org/docs

Installation

composer require phpstats/cmdstanphp

CmdStanPHP needs PHP 8.2 to 8.5, CmdStan 2.37 or later, and a C++ toolchain, and does not need R. Installation covers the requirements, CmdStan, and the optional extensions.

Example

This Stan program estimates the probability theta of a success from N yes-or-no outcomes y. Save it as bernoulli.stan:

data {
  int<lower=0> N;
  array[N] int<lower=0, upper=1> y;
}
parameters {
  real<lower=0, upper=1> theta;
}
model {
  theta ~ beta(1, 1);
  y ~ bernoulli(theta);
}

Compile it, sample four chains, and summarize theta:

use function CmdStan\cmdstan_model;

$model = cmdstan_model(__DIR__.'/bernoulli.stan');

$data = [
    'N' => 10,
    'y' => [0, 1, 0, 0, 0, 0, 0, 0, 0, 1],
];

$fit = $model->sample(
    data: $data,
    seed: 123,
    chains: 4,
    parallel_chains: 4,
    show_messages: false,
);

echo "variable   mean    sd    q5   q95  rhat ess_bulk\n";
foreach ($fit->summary('theta') as $r) {
    printf(
        "%-8s %6.3f %5.3f %5.3f %5.3f %5.3f %8.0f\n",
        $r['variable'],
        $r['mean'],
        $r['sd'],
        $r['q5'],
        $r['q95'],
        $r['rhat'],
        $r['ess_bulk'],
    );
}

variable   mean    sd    q5   q95  rhat ess_bulk
theta     0.257 0.124 0.080 0.485 1.001     1232

Getting started explains the output and the convergence checks.

Documentation

The documentation has guides, worked examples, and Functions and methods, an index of every function and every model and fit method.

Compatibility

CmdStanPHP does not implement every CmdStanR feature. The compatibility page maps CmdStanR's names to CmdStanPHP's and lists the differences and the missing features. Verification reports the comparisons with CmdStanR, posterior, loo, and bayesplot.

Development

composer install
composer test                 # unit tests, no CmdStan needed
composer test:integration     # compiles and runs real models
composer test:compatibility   # compares with R; needs CmdStanR and R packages
composer lint

See Contributing.

License

MIT. The example models in resources/examples/ come from CmdStanR under its BSD 3-Clause license (see the NOTICE there). CmdStan and CmdStanR are separate Stan projects. CmdStanPHP is an independent implementation and not an official Stan interface.