CMP command by Roodman and atanhrho meaning

[Aleksandra Anic]

Dear all,

I am using user-written CMP command (Roodman, 2009). Estimation output for one of the models does not calulate the st. error and correspondingly z-statistic, p-value, confidence interval for atanhrho. Therefore, I do not know if the correlation between equation errors is significant or not. What does it signals? Atanhrho is the arc-tang of corr between the two equations.

Thank you for your help.

Best regards,
Aleksandra

[David Roodman]

A missing standard error usually indicates a problem in the model fitting. The estimate of rho may have converged to +1 or -1, so that atanhrho went to +infinity or -infinity. Or the Hessian of the likelihood may appear not to be negative definite. Anyway, it means that it's having trouble fitting your model to the data. You might explore adding or dropping controls to try to pinpoint the source of trouble.
--David

[Aleksandra Anic]

Dear David,

Thank you for your answer.

Best,
Aleksandra