VECM step by step

[Micheal Wembley]

Hi everyone,

I am doing a time-series analysis using the vecm model.

These are the steps that I have performed:

• ADF test (my variables are stationary at first difference)
• Optimal number of lags (varsoc command in stata)
• Johansen test for cointegration (I have one cointegrating equation)
• VECM estimation results

As diagnostic tests:

• Impulse response function
• Jarque-Brera test for normality
• Stability test (Eigenvalue Stability Condition)

Are the steps correct? Should I add something else? Do I have to add the Granger Causality test?

Thanks in advance!

[Joro Kolev]

The steps are correct, you do not need the test for normality, VARs can be justified on large sample grounds. About the Granger causality, it depends on what is your research question. But once you estimate the VAR, you can see (and test) from the estimated parameters what is Granger causing what.

[FELICITY FORD]

Its correct on my cite either