Market Data Hygiene: Statistical Methods for Detecting Problematic Data

This article has been expanded into a three-part series:

• Part 1: Statistical Methods for Detecting Bad Data: Point anomaly detection (tick tests, NBBO, MAD/IQR, volatility-adjusted thresholds) and systematic error identification (staleness, bid-ask inversion, timestamp drift)

• Part 2: Cross-Validation and Contextual Analysis: Cross-asset validation, time-based patterns, venue-specific considerations, and multi-source triangulation

• Part 3: Reference Data and Historical Integrity: Corporate actions, index membership, point-in-time correctness, and building a validation framework

Start with Part 1.