pyhctsa.operations.correlation.pos_neg_asymmetry¶

pyhctsa.operations.correlation.pos_neg_asymmetry(y)¶

Asymmetry of local dynamics between positive and negative regimes.

Splits the time series by the sign of each value (assumes y is z-scored, so the split is around the mean) and asks whether the one-step-ahead dynamics differ between the two regimes: is the series more volatile, or more persistent (higher one-step autocorrelation), when the current value is above vs. below its mean? This targets a form of distribution-dynamics interaction not captured by stick_angles, which instead compares the distribution of local slopes within each same-sign subsequence.

Parameters:¶

y : array-like¶

The input time series (assumed z-scored: the regime split threshold is 0).

Returns:¶

The conditional volatility and one-step autocorrelation of the positive/negative regimes, and normalized contrasts between them (the volatility contrast is a leverage-effect-style statistic; the autocorrelation contrast is a threshold-AR(1)-style statistic). Also isolates the two zero-crossing transition types (positive-to-negative, negative-to-positive) – posMask/negMask are conditioned on the current value only, so they mix crossing and non-crossing steps together; the crossing-specific fields ask instead whether the jump at a regime switch is itself asymmetric (e.g. sharper downward crossings than upward ones).

Return type:¶

dict