nessai.posterior

Functions related to computing the posterior samples.

Module Contents

Functions

compute_weights(samples, nlive[, expectation])	Returns the log-evidence and log-weights for the log-likelihood samples
draw_posterior_samples(nested_samples[, nlive, n, ...])	Draw posterior samples given the nested samples.

nessai.posterior.compute_weights(samples, nlive, expectation='logt')

Returns the log-evidence and log-weights for the log-likelihood samples assumed to the result of nested sampling with nlive live points

Parameters:

samplesarray_like

Log-likelihood samples.

nliveUnion[int, array_like]

Number of live points used in nested sampling.

expectationstr, {logt, t}

Method used to compute the expectation value for the shrinkage t. Choose between log <t> or <log t>. Defaults to <log t>.

Returns:

float

The computed log-evidence.

array_like

Array of computed weights (already normalised by the log-evidence).

nessai.posterior.draw_posterior_samples(nested_samples, nlive=None, n=None, log_w=None, method='rejection_sampling', return_indices=False, expectation='logt')

Draw posterior samples given the nested samples.

Requires either the posterior weights or the number of live points.

Parameters:

nested_samplesstructured array

Array of nested samples.

nliveint, optional

Number of live points used during nested sampling. Either this arguments or log_w must be specified.

nint, optional

Number of samples to draw. Only used for importance sampling. If not specified, the effective sample size is used instead.

log_warray_like, optional

Array of posterior weights. If specified the weights are not computed and these weights are used instead.

methodstr

Method for drawing the posterior samples. Choose from

• 'rejection_sampling'

• 'multinomial_resampling'

• 'importance_sampling' (same as multinomial)

return_indicesbool

If true return the indices of the accepted samples.

expectationstr, {logt, t}

Method used to compute the expectation value for the shrinkage t. Choose between log <t> or <log t>. Defaults to <log t>. Only used when log_w is not specified.

Returns:

posterior_samplesnumpy.ndarray

Samples from the posterior distribution.

indicesnumpy.ndarray

Indices of the accepted samples in the original nested samples. Only returned if return_indices is True.

Raises:

ValueError

If the chosen method is not a valid method.