Errors are not normally distributed in a reduced form VAR

[Katherine Oleas]

Hello everyone.

I am estimating a VAR in reduced form for the IDEAC, inflation and a financial index. When performing the post estimation tests, it turns out that the null hypothesis of normality of the errors is rejected, since this requirement is not met, does anyone know how I should run my VAR now?
var dtideac dtinflation dtPCA, lags(1/4)

[George Ford]

normality is an issue of the residuals, and even then is not terribly important. the tests almost always reject especially in large samples.

could try log versions of the regressors and see if that helps (if everything is positive in the levels).

could be a structural break.

[Katherine Oleas]

I'm new to stata, maybe you could tell me what is the command to see log versions of the regressors, please?

[George Ford]

just take the log of the variables before differencing.

g lx = ln(x)