Time Series Class

VAR in differences.... not true if cointegration exists
VECM captures the true specification.....better fit

VAR not usually conducted by taking the difference?
or HP filter? VECM in HP filtered series?
interest rate uses raw data? average to apply to quarterly?
serially correlated if averaged?
recover the true monetary policy shock?

trend stationary process.... useful in practice?

Using structural break dummies to recover stationarity?
then not better than sticking to the unit root hypothesis?
(other than forecasting purposes)
Markov switching process as null rather than unit root?