Discuss the concept of vector autoregression (VAR) in multivariate time series analysis.
Instruction: Explain the VAR model, its components, and its application in analyzing systems with multiple interrelated time series.
Updated
Example Answer
A VAR models several time series together. Each equation uses earlier values of that variable and the other variables, plus an innovation. With two variables and two lags, each equation has four lagged predictor terms, so the number of parameters grows quickly as the system gets larger.
I would use a small, justified set of series and check whether a stationary VAR is suitable. A log transformation can stabilize scale-dependent variation but does not automatically remove a stochastic trend. Cointegrated levels may need an error-correction model to retain their long-run relation. After fitting, I would inspect stability and residuals, then compare later-period forecasts with simpler baselines. Predictive lag coefficients alone do not establish causal effects.
Make it your own
Choose two related series and explain which information each equation uses.
Why this works
It makes the joint model concrete and states its assumptions.
Follow-up to practice
How does adding variables change the number of parameters?