qis.unsmooth_returns_ar1_ewma

qis.unsmooth_returns_ar1_ewma(returns, span=20, mean_adj_type=MeanAdjType.EWMA, warmup_period=10, max_value_for_beta=0.75, min_value_for_beta=-0.25, apply_ewma_mean_smoother=True, non_negative=False, non_negative_tol=0.0, insufficient_data=InsufficientData.NAN)[source]

rolling AR(1) unsmoother; a shim over adjust_returns_with_ar with ar_order=1.

Retained for existing callers and for its AR(1) defaults, which differ from the general engine: span=20 and warmup_period=10. New code should call adjust_returns_with_ar directly and state the order.

Parameters:

  • returns (DataFrame) – observed returns of the appraisal-based series, one column per fund

  • span (int) – EWM span of the rolling AR(1) beta estimate

  • mean_adj_type (MeanAdjType) – mean subtracted before estimating the beta; see MeanAdjType

  • warmup_period (int | None) – leading periods blanked while the EWM state converges

  • max_value_for_beta (float | None) – cap on the estimated beta. The inversion divides by 1 - beta, so the cap is what keeps the denominator away from zero; the default 0.75 holds it at or above 0.25

  • min_value_for_beta (float | None) – floor on the estimated beta, below which negative autocorrelation is treated as noise rather than smoothing

  • apply_ewma_mean_smoother (bool) – smooth the estimated beta itself before inverting

  • non_negative (bool) – clip the unsmoothed returns at non_negative_tol

  • non_negative_tol (float) – the clipping level used when non_negative

  • insufficient_data (InsufficientData) – what to do with a column that cannot be identified; see InsufficientData

Returns:

the unsmoothed returns, the estimated betas and the regression R-squared, each in the shape of returns

Return type:

Tuple[DataFrame, DataFrame, DataFrame]