def _sufficient_statistics()

in spotify_confidence/analysis/frequentist/confidence_computers/confidence_computer.py [0:0]
59 lines of code
6 McCabe index (conditional complexity)

    def _sufficient_statistics(self) -> DataFrame:
        if self._sufficient is None:
            kwargs = {
                NUMERATOR: self._numerator,
                NUMERATOR_SUM_OF_SQUARES: self._numerator_sumsq,
                DENOMINATOR: self._denominator,
                BOOTSTRAPS: self._bootstrap_samples_column,
                INTERVAL_SIZE: self._interval_size,
                FEATURE: self._feature,
                FEATURE_SUMSQ: self._feature_ssq,
                FEATURE_CROSS: self._feature_cross,
            }
            groupby = [col for col in [self._method_column, self._metric_column] if col is not None]
            self._sufficient = (
                self._df.groupby(groupby, sort=False, group_keys=True)
                .apply(
                    lambda df: df.assign(
                        **{
                            POINT_ESTIMATE: lambda df: confidence_computers[
                                df[self._method_column].values[0]
                            ].point_estimate(df, **kwargs)
                        }
                    )
                    .assign(
                        **{
                            ORIGINAL_POINT_ESTIMATE: lambda df: (
                                confidence_computers[ZTEST].point_estimate(df, **kwargs)
                                if df[self._method_column].values[0] == ZTESTLINREG
                                else confidence_computers[df[self._method_column].values[0]].point_estimate(
                                    df, **kwargs
                                )
                            )
                        }
                    )
                    .assign(
                        **{
                            VARIANCE: lambda df: confidence_computers[df[self._method_column].values[0]].variance(
                                df, **kwargs
                            )
                        }
                    )
                    .assign(
                        **{
                            ORIGINAL_VARIANCE: lambda df: (
                                confidence_computers[ZTEST].variance(df, **kwargs)
                                if df[self._method_column].values[0] == ZTESTLINREG
                                else confidence_computers[df[self._method_column].values[0]].variance(df, **kwargs)
                            )
                        }
                    )
                    .pipe(
                        lambda df: confidence_computers[df[self._method_column].values[0]].add_point_estimate_ci(
                            df, **kwargs
                        )
                    )
                )
                .pipe(reset_named_indices)
            )
        return self._sufficient