EViews.com

Can someone point me to more detail on what Eviews does when normalizing a series when forecasting? I would like to be able to hand calculate the forecast to check it.

This is from the Eviews help documentation.

If the formula can be normalized, EViews will compute the forecasts of the transformed dependent variable by first forecasting the normalized series and then transforming the forecasts of the normalized series. This methodology has important consequences when the formula includes lagged series. For example, consider the following two models:
series dhs = d(hs)
equation eq1.ls d(hs) c sp
equation eq2.ls dhs c sp

My estimation equation.

equation eq!i_2.ls dlog(hpi_fhfa)-dlog(hpif) dlog(hpi_fhfa(-1))-dlog(hpif(-1)) dlog(hpi_fhfa(-2))-dlog(hpif(-2)) dlog(hpi_fhfa(-3))-dlog(hpif(-3)) dlog(hpi_fhfa(-4))-dlog(hpif(-4)) log(hpi_fhfa(-1))-log(hpif(-1)) d(ur_sa)

My forecast.

eq!i_2.forecast hpa_fcst_!I

Eviews provides forecasts for hpi_fhfa

It just inverts whatever functional form you have.
All you need to calculate Yhat is dlogy_hat (which you get from the equation) and the previous period's Yhat (or Y depending on dynamic or static forecast). For the first period of the forecast you just use the actual value of Y as Yhat.

I'm still not getting it. Can you take a look at my formula (column W) in the attached excel sheet?

I think your algebra is wrong. You can't take the exp of both sides when you have two log terms like that.

Can you tell me what I need to change in the formula in column W?

=EXP(coef!$D$3*((E144/E143)-(L144/L143)))*E144

That's outside the scope of what we here at EViews will provide. But perhaps some other forum user can help.

Ok, how would I do it in Eviews without using forecast?

Same way, you'd invert the function.

Figured it out. Thanks for all your great help Gareth. Great to see nothing has changed around here.

Here is the excel formula for anyone who is interested.

=L145*(E144/L144)*EXP((coef!$D$3*forecast!M145)+(coef!$E$3*forecast!N145)+(coef!$F$3*forecast!O145)+(coef!$G$3*forecast!P145)+(coef!$H$3*forecast!Q145)+(coef!$I$3*forecast!R145))