BayesianBasisExpansionTimeSeries.build#

BayesianBasisExpansionTimeSeries.build(X, y, coords=None)[source]#

Construct the graph and record the time-feature data nodes.

Parameters:
  • X (DataArray) – Input features with dims [“obs_ind”, “coeffs”]; obs_ind must be datetimes.

  • y (DataArray) – Target variable with dims [“obs_ind”, “treated_units”].

  • coords (dict[str, Any] | None) – Coordinates dictionary; must contain “datetime_index”.

Return type:

None

Notes

The trend/seasonality inputs live under their own pm.Data names, so they are added to _build_data_nodes for _rearm_fit_data().