resample_posterior

params.resample_posterior(source, *, n, weight='weight', seed=None)

Resample a weighted parameter table into an unweighted draw matrix.

Rows are drawn with replacement with probability proportional to the weight column, jointly (whole rows), so any correlation in the posterior survives. The weight column is dropped from the result, which carries a fresh RangeIndex named "iteration".

Resampling to an n larger than the input adds no information; it only smooths Monte Carlo noise in downstream expectations.

Parameters

Name	Type	Description	Default
source	pd.DataFrame | str | Path	A weighted parameter table DataFrame, or a path to a CSV file of one, with one column of weights and the rest parameters.	required
n	int	Number of rows in the resampled draw matrix.	required
weight	str	Name of the weight column. Weights must be non-negative and not all zero; they are normalised internally.	'weight'
seed	int | np.random.Generator | None	Integer seed or numpy Generator for the resampling.	None

Returns

Name	Type	Description
	pd.DataFrame	DataFrame with n rows, a RangeIndex named "iteration", and
	pd.DataFrame	one column per parameter (the weight column removed).

Raises

Name	Type	Description
	ValueError	If the table is empty, weight names a missing column, n is not positive, or the weights are negative or sum to zero.

Example

import pandas as pd from heormodel.params import resample_posterior post = pd.DataFrame({“beta”: [0.1, 0.2, 0.3], “weight”: [1.0, 2.0, 1.0]}) resample_posterior(post, n=1000, seed=0).shape (1000, 1)