growth_error_variance

growth_error_variance(
    estimates,
    as_of,
    source,
    *,
    minimum_observations=defaults.MINIMUM_OBSERVATIONS,
    winsorize_fraction=None,
)

Compute one source’s growth track record ahead of an entity’s history.

The variance is the mean of squared errors; it is not a centred (mean-subtracted) variance, since a source with a consistent bias should read as imprecise, not merely as scattered around its own bias.

Parameters

Name	Type	Description	Default
estimates	pd.DataFrame	Long input panel.	required
as_of	pd.Timestamp	The moment to read the panel at.	required
source	str	The source name whose track record is being measured.	required
minimum_observations	int	The fewest growth errors that make the variance meaningful.	defaults.MINIMUM_OBSERVATIONS
winsorize_fraction	float | None	Share of the upper tail of error magnitudes clipped, each history on its own bounds. None clips nothing.	None

Returns

Name	Type	Description
	pd.DataFrame	One row per subject present in the panel as of as_of, carrying f”{source}_growth_error_variance” and f”{source}_growth_error_count”. The variance is NaN below the minimum observation count; the count is always the true one.