Estimating a VEC

[womatar]

Hello Guys,

I have recently picked up time series analysis, and I wanted to make sure that what I am doing is appropriate when estimating a VEC.

1.) I initially test to make sure that all series are integrated of the same order. Once I do, and their order of integration is above 0,
2.) I test for cointegration using the Johansen test. I test using different lag specifications starting with 0 to some value, n. n will depend on my sample size and number of variables.
3.) For those specifications (lags, trend, and intercept considered depending on problem) that pass the necessary criteria for cointegration, I estimate the associated VEC.
4.) I then test the residuals of the VEC models and make further filtering of the specifications based on their normality, serial correlation, and heteroskedasciticty.
5.) If I have more than one specification remaining that pass all my previous steps, I choose the model that minimizes the AIC and SIC. I also make sure the adjustment coefficients are viable and consider weak exogeniety of variables.

Please advise if this approach is okay. This has always resulted in physically/economically sound results.

Thank you.