BayesianPosterior

BayesianPosterior(
    forecast_source,
    settings=BAYESIAN_DEFAULTS,
    consensus_source=CONSENSUS_SOURCE,
)

Prior, model and consensus, blended by their own precision.

Stateless rather than fitted: every precision is computed per subject from that subject’s own point-in-time history, which history supplies as columns on the frame. Nothing is learned across subjects, so fit has nothing to do but verify the frame and hand back a fitted object; the per-subject arithmetic lives entirely in apply.

Attributes

Name	Type	Description
forecast_source	str	The source blended alongside consensus.
settings	BayesianSettings	The arguments to apply under.
consensus_source	str	The source treated as consensus.

Methods

Name	Description
fit	Verify the sources this method needs are present, and learn nothing.

fit

fit(subjects)

Verify the sources this method needs are present, and learn nothing.

Parameters

Name	Type	Description	Default
subjects	pd.DataFrame	Training rows, read only to check that the sources this method needs are present.	required

Returns

Name	Type	Description
	FittedBayesianPosterior	The fitted method, carrying the settings it was constructed with.

Raises

Name	Type	Description
	PanelError	A source this method reads is absent from the frame.