MMMSummaryFactory.prior_predictive#

MMMSummaryFactory.prior_predictive(hdi_probs=None, frequency=None, original_scale=True, output_format=None)[source]#

Create prior predictive summary DataFrame.

Mirrors posterior_predictive() but draws from the prior predictive distribution.

Parameters:
hdi_probssequence of float, optional

HDI probability levels (default: uses factory default)

frequencyFrequency, optional

Time aggregation period (default: None, no aggregation)

original_scalebool, default True

If True, summarize y_original_scale from idata.prior. If False, summarize y from idata.prior_predictive.

output_format{“pandas”, “polars”}, optional

Output DataFrame format (default: uses factory default)

Returns:
pd.DataFrame or pl.DataFrame

Summary DataFrame with columns:

  • date: Time index

  • mean, median: Prior predictive point estimates

  • observed: Observed target values

  • abs_error_{prob}_lower/upper: HDI bounds for each prob