meridian.analysis.review.checks.PotentialBiasCheck

A check for correlation between paid channels and control variables to flag potential confounding.

Inherits From: BaseCheck

meridian.analysis.review.checks.PotentialBiasCheck(
    *,
    meridian: (Any | None) = None,
    model_context: (meridian.model.context.ModelContext | None) = None,
    inference_data: (az.InferenceData | None) = None,
    analyzer: meridian.analysis.analyzer.Analyzer,
    config: meridian.analysis.review.checks.ConfigType,
    selected_geos: (Sequence[str] | None) = None,
    selected_times: (Sequence[str] | None) = None
)

Methods

is_relevant

View source

@classmethod
is_relevant(
    model_context: meridian.model.context.ModelContext,
    inference_data: (az.InferenceData | None) = None
) -> bool

Checks if the check is relevant for the given model context.

run

View source

run() -> meridian.analysis.review.results.PotentialBiasCheckResult

Runs the potential bias check.

This check computes the Pearson correlation between each paid channel's media/RF data and all control variables. It flags channels where the maximum absolute correlation with any control variable exceeds a predefined threshold.

Returns
A results.PotentialBiasCheckResult object containing the results of the check.