bayespecon.diagnostics.bayesfactor.post_prob¶

bayespecon.diagnostics.bayesfactor.post_prob(logml_list, model_names=None, prior_prob=None)[source]¶

Compute posterior model probabilities from marginal likelihoods.

Following the R bridgesampling package’s post_prob() function.

Parameters:¶

logml_list : array-like¶

Log marginal likelihoods, one per model.

model_names : list of str, optional¶

Labels for each model. If None, models are labeled by index.

prior_prob : array-like or None¶

Prior model probabilities. If None, uniform priors are used (all models equally likely a priori).

Returns:¶

Posterior model probabilities (sum to 1), indexed by model names.

Return type:¶

pandas.Series

Examples

>>> post_prob([-20.8, -18.0, -19.0], model_names=["H0", "H1", "H2"])
H0    0.05...
H1    0.72...
H2    0.22...
dtype: float64