Skip to content

K-Dense-AI/scientific-agent-skills / skills/pymc

Builds and checks Bayesian models with PyMC, including hierarchical models, NUTS MCMC, variational inference, mutable-data predictions, posterior predictive checks, diagnostics, and PSIS-LOO model comparison. Use for probabilistic modeling and uncertainty inference in PyMC.

assets/hierarchical_model_template.py

View on GitHub
"""Synthetic hierarchical regression with existing- and new-group predictions.

Targets PyMC 6.3.2 / ArviZ 1.3. This model assumes independent group intercept
and slope effects. For correlated effects use an LKJ Cholesky model instead.
Short runs exercise APIs, not scientific validation.
"""

import argparse
from pathlib import Path

import arviz as az
import matplotlib.pyplot as plt
import numpy as np
import pymc as pm


def build_model(x, groups, y, group_names):
    with pm.Model(coords={'groups': group_names, 'obs': np.arange(len(y))}) as model:
        x_data = pm.Data('X_data', x, dims='obs')
        groups_data = pm.Data('groups_data', groups, dims='obs')
        mu_alpha = pm.Normal('mu_alpha', mu=0, sigma=5)
        sigma_alpha = pm.HalfNormal('sigma_alpha', sigma=2)
        mu_beta = pm.Normal('mu_beta', mu=0, sigma=2)
        sigma_beta = pm.HalfNormal('sigma_beta', sigma=1)
        alpha_offset = pm.Normal('alpha_offset', 0, 1, dims='groups')
        beta_offset = pm.Normal('beta_offset', 0, 1, dims='groups')
        alpha = pm.Deterministic('alpha', mu_alpha + sigma_alpha * alpha_offset, dims='groups')
        beta = pm.Deterministic('beta', mu_beta + sigma_beta * beta_offset, dims='groups')
        sigma = pm.HalfNormal('sigma', 1)
        mu = alpha[groups_data] + beta[groups_data] * x_data
        pm.Normal('y_obs', mu, sigma=sigma, observed=y, shape=x_data.shape[0], dims='obs')
    return model


def predict_new_group(idata, x_new, random_seed=42):
    """Predict several observations sharing ONE previously unseen group.

    Draw new group effects once per posterior draw and reuse across x_new;
    add independent residual noise per observation. Returns (mean, outcome)
    arrays with shape (chain, draw, new_observation).
    """
    x_new = np.asarray(x_new, dtype=float)
    if x_new.ndim != 1 or not np.isfinite(x_new).all():
        raise ValueError("x_new must be a finite vector")
    rng = np.random.default_rng(random_seed)
    posterior = idata.posterior
    mu_a, mu_b = posterior['mu_alpha'].values, posterior['mu_beta'].values
    alpha_new = mu_a + posterior['sigma_alpha'].values * rng.normal(size=mu_a.shape)
    beta_new = mu_b + posterior['sigma_beta'].values * rng.normal(size=mu_b.shape)
    mean = alpha_new[..., None] + beta_new[..., None] * x_new
    outcome = mean + posterior['sigma'].values[..., None] * rng.normal(size=mean.shape)
    return mean, outcome


def run(draws=1000, tune=1000, chains=4, output_dir='hierarchical_results'):
    rng = np.random.default_rng(42)
    n_groups, n_per_group = 6, 12
    group_names = [f'group_{i}' for i in range(n_groups)]
    groups = np.repeat(np.arange(n_groups), n_per_group)
    x = rng.normal(size=len(groups))
    alpha = rng.normal(1, .6, n_groups)
    beta = rng.normal(1.5, .3, n_groups)
    y = alpha[groups] + beta[groups] * x + rng.normal(0, .5, len(x))
    model = build_model(x, groups, y, group_names)
    output = Path(output_dir)
    output.mkdir(parents=True, exist_ok=True)

    def save(collection, name):
        collection.savefig(output / name, dpi=150, bbox_inches='tight')
        plt.close(collection.viz['figure'].item())

    with model:
        prior = pm.sample_prior_predictive(draws=100, random_seed=40)
    save(az.plot_ppc_dist(prior, group='prior_predictive', num_samples=30), 'prior_predictive.png')
    # Inspect prior implications before fitting real data.
    with model:
        idata = pm.sample(draws=draws, tune=tune, chains=chains, cores=1,
                          nuts_sampler='pymc', target_accept=.95, random_seed=42,
                          progressbar=False)
        pm.compute_log_likelihood(idata, progressbar=False)
        pm.sample_posterior_predictive(idata, extend_inferencedata=True,
                                      random_seed=43, progressbar=False)
    variables = ['mu_alpha', 'sigma_alpha', 'mu_beta', 'sigma_beta', 'sigma']
    summary = az.summary(idata, var_names=variables + ['alpha', 'beta'],
                         round_to='none', ci_kind='hdi', ci_prob=.95)
    print(summary)
    print('[INFO] Inspect diagnostics and prior sensitivity; six groups weakly constrain group scales.')
    save(az.plot_trace_dist(idata, var_names=variables), 'trace.png')
    save(az.plot_ppc_dist(idata, num_samples=30), 'posterior_predictive.png')
    save(az.plot_forest(idata, var_names=['alpha', 'beta'], combined=True,
                        ci_kind='hdi', ci_probs=[.5, .95]), 'group_parameters.png')
    new_x, new_groups = np.array([-1., 0., 1.]), np.array([0, 2, 4])
    with model:
        pm.set_data({'X_data': new_x, 'groups_data': new_groups}, coords={'obs': np.arange(3)})
        predictions = pm.sample_posterior_predictive(idata, predictions=True,
                                                   random_seed=44, progressbar=False)
        pm.set_data({'X_data': x, 'groups_data': groups}, coords={'obs': np.arange(len(y))})
    interval = az.hdi(predictions.predictions['y_obs'], prob=.95)
    interval.to_dataframe(name='prediction_interval').to_csv(output / 'existing_group_intervals.csv')
    new_mean, new_outcome = predict_new_group(idata, new_x)
    np.savez(output / 'new_group_predictions.npz', x=new_x, mean=new_mean, outcome=new_outcome)
    idata.to_netcdf(output / 'posterior.nc')
    predictions.to_netcdf(output / 'predictions.nc')
    summary.to_csv(output / 'summary.csv')
    return idata, predictions, model


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument('--draws', type=int, default=1000)
    parser.add_argument('--tune', type=int, default=1000)
    parser.add_argument('--chains', type=int, default=4)
    parser.add_argument('--output-dir', default='hierarchical_results')
    run(**vars(parser.parse_args()))
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall