lab module
Watch a Kalman filter track a noisy signal
A one dimensional filter estimates a hidden price from noisy observations. Move Q and R to see how model trust changes the estimate and uncertainty.
hidden statenoisy observationsfiltered estimateone sigma band
Increasing Q tells the filter the hidden state can move more between observations, so the estimate reacts faster and the band widens. Increasing R tells the filter that observations are noisy, so it trusts the model more and moves more slowly.
project link
This demo is a small teaching version of the filtering ideas used in the Bayesian Market Filters project.