Re: st: Use of statsby in var command

["Brian P. Poi" <bpoi@stata.com>]

On Thu, 18 Sep 2008, J�rn van Halteren wrote:

> The STATA manual says that vector autoregression (var) works with the
> command "statsby". However, I always get the error r(451) "repeated time
> values in sample an error occurred when statsby executed var". If I run the
> command with simple by it works, but I need the stats.
>
> Can someone help me?
To use -var- and other time-series commands with panel data, you need to 
use the basepop() option of -statsby-:

   . webuse grunfeld, clear
   . statsby _b, by(company) basepop(company == 1): var i m, lags(1/2)


By default, -statsby- first tries to fit the model to the entire dataset 
to check command syntax.  Commands like -var- for time-series data will 
complain when used with panel data.  The basepop(company == 1) option 
tells -statsby- to check syntax by using only the first panel's 
observations, which constitute a time series that -var- can handle.

   -- Brian Poi
   -- bpoi@stata.com