Simula@BI: Realized principal component analysis of noisy high-frequency data

We propose a pre-averaging version of the realized principal component analysis (RPCA) estimator, which is the extension of the classical principal component analysis to high-frequency data. In practice, tick-by-tick data on asset prices are contaminated by measurement error due to microstructural noise.

We suggest aconsistent noise-robust estimator of the spot covariance matrix employing a pre-averaging technique. Then, building on recent theory about volatility functional estimation we derive the realized eigenvalue, eigenvector and principal component estimators for the noisy setting. We inspect the accuracy of the new estimator with simulation results, while an empirical application shows how it works in practice.

• Read more about the speaker. 

• Visit our website to read more about the research center Simula@BI