environmentaltools.temporal.inverse_transform

environmentaltools.temporal.inverse_transform(data: DataFrame, params: dict, ensemble: bool = False)[source]

Reverse power transformation to original data scale.

Applies inverse of Box-Cox or Yeo-Johnson transformation to return data to its original scale after analysis in transformed space.

Parameters:

  • data (pd.DataFrame) – Transformed data to convert back to original scale

  • params (dict) –

    Parameter dictionary containing transformation information:

    • If ensemble=False: uses params[‘transform’][‘method’] and params[‘transform’][‘lambda’]

    • If ensemble=True: uses params[‘method_ensemble’] and params[‘lambda_ensemble’]

  • ensemble (bool, default=False) – Whether to use ensemble transformation parameters instead of standard transformation parameters

Returns:

Data in original scale with same index as input

Return type:

pd.DataFrame

Notes

The inverse transformations:

  • Box-Cox: x = (λy + 1)^(1/λ) for λ ≠ 0

  • Yeo-Johnson: Piecewise inverse depending on sign and λ

This function must use the same lambda value that was used in the forward transform() call to ensure correct inversion.

See also

transform

Forward power transformation

sklearn.preprocessing.PowerTransformer

Underlying implementation

Examples

>>> # After transforming and analyzing
>>> data_original = inverse_transform(data_transformed, params)
>>> # For ensemble models
>>> data_original = inverse_transform(data_ensemble, params, ensemble=True)