The RATS Software Forum

Posted: **Fri Jun 04, 2021 4:08 pm**

Hi Tom,

I am generating h-steps (e.g. h=4) ahead forecasts within recursive regressions and want to save only the fourth forecast and the associated stderrs step, from each recursive regression. I am comparing FORECAST vs. UFORECAST.

Code: Select all

do regend = ibegin, iend
   boxjenk(print,constant,ar=||1||,define=foreeq_fxRate,maxl) fxRate istart regend resids_fxRate
*
   forecast(skip=(4-1),from=regend+1,steps=4,stderrs=stderrs_D4_fxRate) 1
   # foreeq_fxRate yhat_D4_fxRate
*
   uforecast(equation=foreeq_fxRate,from=regend+1,stderrs=u_stderrs_D4_fxRate) u_yhat_D4_fxRate regend+1 regend+4
   compute fcst_D4_f(regend+4) = u_yhat_D4_fxRate(regend+4)
   compute fcst_D4_SD(regend+4) = u_stderrs_D4_fxRate(regend+4)
end do regend

The h=4 step ahead forecasts: yhat_D4_fxRate (FORECAST) and fcst_D4_f (computed via UFORECAST) are the same.

The stderrs: stderrs_D4_fxRate (FORECAST) and u_stderrs_D4_fxRate (UFORECAST) are the same, but not for h=4.

I want the h=4 stderrs, which are computed in fcst_D4_SD. Correct?

Why are the h=4 stderrs not in stderrs_D4_fxRate (FORECAST), having used skip=(4-1)?

thanks,
Amarjit

Posted: **Mon Jun 07, 2021 10:37 am**

SKIPSAVE doesn't apply to the STDERRS calculations (only to the RESULTS and ERRORS). I don't believe anyone else has ever requested it. After all, the forecasts and the errors depend on the data, but the standard errors only depend upon the model and won't change much with a slightly different sample.

If you really want to do that, use the first method on

https://estima.com/docs/RATS%2010%20Use ... f#page=186

(that is, put the STDERRS into one series and pull out the step 4 value into a separate one).

The RATS Software Forum

The hth-step ahead forecast and associated stderrs

The hth-step ahead forecast and associated stderrs

Re: The hth-step ahead forecast and associated stderrs