Limit obersvations used in a regression

[Chris Ostriker]

Hello, in order to find the optimal lag length of my VAR-Model I want to include one lag at a time iterativly and compare them with the likelihood ratio test. However, I have to run the one-lag-VAR model with one less observation to make it comparable to the VAR model with 2 lags. How do I restrict the number of observations used in the regression? (Without introducing a new variable)
Thanks

[Clyde Schechter]

Although I do not myself use -var-, on the assumption that it works like any other Stata estimation command, just work backwards. The sample with the largest number of lags will be the most restricted of all. Start with the largest number of lags you would be willing to consider, run that model first. Then run:

gen byte included = e(sample)

Then iteratively reduce the number of lags one at a time, and each time in your -var- command impose the restriction -if included-.

[Joro Kolev]

Also check out

varsoc -- Obtain lag-order selection statistics for VARs and VECMs