Get started

This page takes you from installation to a fitted model you can read: a small regression, its diagnostics, and a prediction with uncertainty. It assumes you know what a prior and a posterior are, and nothing about rustmc. Ten minutes.

Install the Python package with pip install rustmc. NumPy is required; ArviZ and Matplotlib are optional through pip install "rustmc[viz]".

Estimate an instrument's gain

import numpy as np
import rustmc as rmc

rng = np.random.default_rng(42)
x = np.linspace(-2, 2, 100)
y = 0.3 + 1.2 * x + rng.normal(0, 0.2, len(x))

model = rmc.ModelBuilder()
offset = model.normal_prior("offset", 0.0, 1.0)
gain = model.normal_prior("gain", 1.0, 0.5)
noise = model.half_normal_prior("noise", 0.5)
model.normal_likelihood("reading", offset + gain * "x", noise, "y")
compiled = model.compile()
fit = compiled.sample(
    {"x": x, "y": y}, chains=4, warmup=1000, draws=1000,
    seed=42, show_progress=False,
)
print(fit.summary())

The posterior estimates the offset, gain, and measurement noise together. Choose priors in the units of your measurements.

Inspect R-hat, bulk/tail effective sample size, Monte Carlo error, and divergences. R-hat near one and enough effective samples are necessary checks, not proof that the model describes your data. Compare predictive draws with observations too.

prediction = fit.predict({"x": np.array([-1.0, 0.0, 1.0])}, seed=43)
interval = np.quantile(prediction["reading"], [0.025, 0.975], axis=(0, 1))
print(interval)

Draws have (chain, draw, observation) axes. expected=True returns conditional means; the default also samples observation noise.

Reuse compiled for another instrument. Use a batch for independent fits or one partial-pooling model for related groups. See custom models for the full construction API.

For source builds, follow the repository's contributing guide.