Graphical causal models for time
series econometrics: some recent
developments and applications
Alessio
Moneta, Max Planck Institute for Economics, Jena, Germany
Structural vector-autoregressive models are potentially very useful
tools for guiding economic policy. I present a recently developed method
to estimate and identify the causal structure underlying the data
generating process. The method, which is based on graphical models,
exploits conditional independence tests among estimated VAR residuals to
infer the causal relationships among contemporaneous variables. I first
show how this method works in the Gaussian linear setting. Then I
present some developments for both the linear non-Gaussian and nonlinear
settings.
[NIPS 2009 Causality and Time Series
Mini-Symposium]