Kalman Filters: Fitting a model instead of a line
By Client Solutions & Research at Sanlam Investments
Executive Summary
Kalman filtering is presented as an adaptive and rather accurate method for tracking benchmarks and funds and analysing constituents in asset management applications such as portfolio construction. The paper explains why the underlying model is a far more realistic model of how fund and index constituents move over time than the model on which linear regression is based. The description of Kalman filtering is entirely non-mathematical and intuitive. Kalman filtering is compared and contrasted with linear regression. Situations in which Kalman filtering is superior are explained and two practical applications in a South African context are provided. The paper concludes by pointing to other applications of Kalman filters in asset managements and future projects.
Comments are closed.