Vector autoregressive models with measurement errors

Alexandre Galvão Patriota

Abstract: This presentation introduces a new approach for the estimation of vector autoregressive (VAR) models in the case of multivariate time series with measurement errors. Consistent parameter estimators and their asymptotic distributions are presented. The estimation procedure does not require any iterative process. The theoretical results and measurement error effects on parameter estimates were evaluated by using computational simulations. The simulation results show that the proposed approach produces test size close to the adopted nominal level (even for small samples) and has a good performance around the null hypothesis when the time series are subject to small and moderate measurement errors. The applicability and usefulness of the proposed approach are illustrated using a functional magnetic resonance imaging dataset.

This is a joint work with Betsabé Blas and João Sato.