Meaning
Persistent periods of volatility characterize the fluctuations in a time series where the standard deviation remains within a specific range before shifting to another level. Identifying a variance regime allows a founder to understand if recent swings in metrics are normal or indicative of a structural shift. This classification is ordinal, often moving between various distinct states.
The boundary of a regime is defined by the stability of the dispersion.
Level Persistence
Temporal duration of a specific volatility state indicates the underlying health of the process. A variance regime typically lasts for a sustained number of observations. This persistence allows for the prediction of future noise levels.
Short bursts of volatility that do not last are treated as outliers rather than regime changes.
Transition Trigger
External or internal events cause the move from one level of variance to another. In a variance regime model, the probability of a shift is monitored closely. A sudden increase in uncertainty often precedes a change in the mean of the data.
Identifying these triggers early provides a warning of systemic instability.
Signal Amplitude
Magnitude of the fluctuations defines the operational limits of the system. Each variance regime requires a different sensitivity setting for alerts and monitors. In a high-variance state, small changes are ignored to avoid false positives.
Conversely, a low-variance state allows for the detection of very subtle signals.
