Online Recursive Estimation for Streaming Market Data

Summary

This Pine library collects streaming methods for processing market observations incrementally, with many updates designed to use constant work per observation. Its stated tools cover recursive filters and extrema, rolling sums, quantile and expectile estimates, tail means, robust location and scale, covariance, correlation, regression, and heavy-tail distribution estimates. It also defines transformations for relative returns and projections, along with market participation and dispersion models.

The design separates stateful estimators from ordinary mathematical transforms and specifies how missing observations and undefined domains are handled. Some invalid configurations raise runtime errors, while unavailable inputs yield missing values for the caller to manage. The included demonstration plots a recursive center and dispersion bands, with diagnostics such as innovation and standardized innovation. The document describes implementation scope and conventions rather than a trading strategy; it reports no empirical comparison, forecasting accuracy, or trading performance. The usefulness of any estimator depends on its configuration, inputs, and validation for the intended market data.

Key ideas

• The library updates many statistical estimators incrementally for streaming observations.
• Its methods include robust location and scale, tail statistics, dependence measures, and regression.
• Market participation and dispersion models let callers select different reference domains and evidence types.
• Missing data and undefined calculations remain missing so the caller can choose a fallback policy.
• The demonstration illustrates recursive equilibrium and dispersion bands without reporting trading results.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.