bayesplain.regression

bayesplain.regression(x, y, prior='conventional', labels=None, unit='', reference=0.0, n_draws=None, seed='unset', data=None)[source]

Estimate how much y changes per unit of x, with one predictor.

Parameters:
x : array_like or str

Predictor and outcome, or the names of two columns in data. Rows where either is missing are dropped, and the summary says how many.

y : array_like or str

Predictor and outcome, or the names of two columns in data. Rows where either is missing are dropped, and the summary says how many.

prior : str, float, or EffectSizePrior, default 'conventional'

Cauchy prior on the standardised slope. Affects the Bayes factor only; the posterior and credible interval do not depend on it.

labels : sequence of str, optional

Names of the predictor and the outcome, used in output.

unit : str, optional

Unit of the slope, e.g. "dollars per square foot".

reference : float, default 0.0

The slope to compare against. Zero means “no relationship”; set a policy-relevant value when one exists.

n_draws : int, optional

Number of draws. Defaults to the package setting.

seed : int or None, optional

Seed for the draws.

data : data frame, optional

Where to look x and y up when they are column names. The column names become the labels unless labels is given.

Returns:

Call .summary(), .predict(x=...) for a forecast at a new value of the predictor, and .plot(kind='line') for the fitted line with its credible band.

Return type:

Result

Examples

>>> import numpy as np, bayesplain as bp
>>> rng = np.random.default_rng(0)
>>> sqft = rng.uniform(800, 3000, 60)
>>> value = 150 * sqft + rng.normal(0, 60_000, 60)
>>> res = bp.regression(sqft, value, labels=["floor area", "assessed value"],
...                     unit="dollars per sq ft")
>>> res.exact
True
>>> lo, hi = res.interval(kind="eti")
>>> bool(np.allclose((lo, hi), res.frequentist.interval))
True