Title: The expectation maximization algorithm for the state space model with correlated errors
Authors: Javier Cara - Universidad Politecnica de Madrid (Spain) [presenting]
Abstract: The state space model is a well known and used time series model, specially when dealing with multivariate time series. In practice, this model can be estimated using the Expectation Maximization algorithm assuming both the error in the state equation and the error in the observation equation are not mutually correlated. However, there are some situations where this hypothesis is not valid. The equations for estimating the state space model with correlated errors using the Expectation Maximization algorithm are presented. Finally, these equations are applied to time series of acceleration data measured in civil engineering structures, one example of multivariate state space model with correlated errors.