Re: [Gretl-devel] Forecast based on ordered probit
Am 16.12.2015 um 17:22 schrieb Artur Tarassow:
Thanks for the response, Sven. $fcast provides (as expected) the
same
information.
The point is that I need for the computation of the out-of-sample
probabilities A) the cut-values (provided via $coeff) as well as the
forecasted "z-hat values" which I am not aware of on how to access these.
I don't have much experience (=none?) with these type of models, but
from looking at section 32.2 of the user guide it would seem to me:
a) \hat{\beta} is available via $coeff
b) The X-es are available as observables for forecasting.
c) So you can get \hat{z} = X \hat{\beta}
d) The alphas (cuts) are also available in $coeff
And that's all you need, no? But maybe I'm missing something.
cheers,
sven