VECM Stages and Interpretation

[jonsie92]

Firstly I hope this post is in the correct location and I apologies if it is not.

I am trying to compute the effects of changes in money demand on the USD/GBP exchange rate. As I understand it if my variables are integrated of order 1 I(1) but at least one cointegrating relationship exists then the VECM should be the correct choice of model, after much reading I have been unable to find any exact step by step procedure to this matter and am therefore hoping that someone can point me in the direction of some resources or that together we can define the order of steps.

So far my following process is

1. Conduct Unit Root test: To ensure that all variables are I(1), as much Macroeconomic data is.
2. Conduct Johansen Cointegration test: To determine the number of cointegrating relationships
3. Run VECM: Using the result of 2. we can now run a VECM - I am unsure of how many lags to include here is there a formal test such as the AIC or BIC that can be run to determine the number of lags.
4. Impulse Response Functions:
5. Forecasting: Using out of sample data.

I also think there should be a granger causality test but i'm not sure where in the order of processes this should come. Does anybody know if these steps are correct and in the correct order. Additionally I have seen suggestions of using "lag selection criteria" after running "unrestricted VAR", to establish the number of lags to include, is this advisable or should I use 4 Lags to reflect my data being observed quaterly