On Wed, 30 May 2012, mastro2387(a)tiscali.it wrote: > I have a problem with running a VAR: i selected three > endogenous variables and one dummy as exogenous, 8 lags > chosen by lag selection option, with constant, trend and > robust error options selected, but gretl is saying to me > matrix is no positive definite: how can I sort this issue > out?
That message means that your set of regressors is close to perfect collinearity -- not so close that OLS estimates cannot be obtained, but close enough that computation of robust standard errors fails. The error message could be improved. Allin Cottrell
