Dear all

I am using VECM for the first time so please apologise me if my questions are so basic.
I have 4 variables: two are change variables and two are level variables.
I have first used varsoc to determine the number of lags and 2 lags are recommended.
I then used the two lags in vecrank to determine the number of cointegrated equations. The rank recommends 3 cointegrated equations.

Therefore, I estimated my vec with 2 lags and 3 cointegrated equations using the options lag(2) and rank(3).

My questions are:

1- Is it reasonable to have two lags and 3 ranks? is there any limitation of the number of ranks being less than the number of lags in the vecm?

2- I am not sure what trend option I should use given that two variables are are first-differenced (change) and two are in levels. For example trend(rtrend) or trend(rconstast)..?

3- Just a small background, I have already estimated a VAR with the four variables in first differences but I am now using vec, keeping two series in first-difference as they were originally, and using the levels of the other two. The reason for doing that is my interest in also examining the impact of shocks in the first-differenced series on the "levels" of the other two series using IRF.
is this approach acceptable?


Thanks