The other option is, first, to apply the within-transformation to the data, then, to use the 'var' command as above but without exog().īun, M.

Xi: var y1 y2 圓, exog(i.id_group) lag(1/3)ĭo you think this makes the trick? Or, do you believe that this possibly generates some kind of error in the estimation?Īny suggestion, warning, advice is very much appreciated. Then, to perform the VAR estimation with the Least Squares Dummy Variables estimator, I use the 'var' command and insert dummies as exogenous variables for each longitudinal unit: Nonetheless, I made it work by disregarding the panel structure and by considering the data as long time series. The 'var' command in Stata does not work because of the panel structure. However, I would like to enjoy the flexibility of the 'var' built-in command available in Stata, and I come to my question now. Then, this is appropriate for large-N/short-T panels, which is not my case.Ģ) xtvar, which instead applies the Least Squares Dummy Variables Corrected estimator à-la Bun and Kiviet (2006). Although, there are two user-written routines available on the internet: 1) pvar, 2) xtvar.ġ) pvar (by I.Love) applies a GMM estimator à-la Arellano+Bond to deal with Nickell's bias. For an application: Lof+Malinen-2014.Īs we know, there are no built-in commands to estimate panel-VAR in Stata. Then, as stated in Canova+Ciccarelli-2013 and Canova-2007, the Least Squares Dummy Variables estimator should be a sound estimator to apply. Given the dimension of my panel (long T, short N), Nickell's bias should be negligible. I have a macro panel (N=8/9, T=65) and I want to estimate a VAR including three endogenous variables. I have read the previous posts about this topic, then this should not be a repetition. This is about the estimation of a panel VAR.

Stata 13 for dummies