This paper presents the MARX package for the analysis of mixed causal-noncausal autoregressive processes with possibly exogenous regressors. The distinctive feature of MARX models is that they abandon the Gaussianity assumption on the error term. This deviation from the Box-Jenkins approach allows researchers to distinguish backward (causal) and forward-looking (noncausal) stationary behavior in time series (see e.g. Hecq et al., 2016, for an overview). The MARX package offers functions to simulate, estimate and select mixed causal-noncausal autoregressive models, possibly including exogenous regressors. The procedures for this are discussed in Hecq et al. (2016) for the MAR, and Hecq et al. (2017) for the MARX respectively.
|Number of pages||18|
|Publication status||Published - 2017|
|Series||Social Science Research Network|
- c22 - "Single Equation Models; Single Variables: Time-Series Models; Dynamic Quantile Regressions; Dynamic Treatment Effect Models"
- e31 - "Price Level; Inflation; Deflation"
- e37 - Prices, Business Fluctuations, and Cycles: Forecasting and Simulation: Models and Applications
- mixed causal-noncausal autoregressive process
- model selection
Hecq, A., Lieb, L., & Telg, S. (2017). Simulation, Estimation and Selection of Mixed Causal-Noncausal Autoregressive Models: The MARX Package. SSRN. Social Science Research Network https://ssrn.com/abstract=3015797