EViews.com

Posted: **Mon Jul 09, 2018 9:23 am**

I'm stuck on this. I am using EViews 10+ Enterprise Edition June 22 2018.

This is a simple loop for ETS estimation:

%towers2 = "CYVR CYEG CYYC CYXE CYWG CYYZ CYTZ CYOW CYUL CYHZ CYXX CZBB CYBW CYRC CZVL CYFC CYMM CYQX CYHM CYLW CYKF CYNJ CYXU CYQM CYHU CYOO CYPK CYXS CYQB CYQR CYJN CYAM CYYT CYQT CYKZ CYYJ CYXY CYQG CYAV CYZF CYHC"
%flight_rule2 = "IFR VFR"

for %series1 {%flight_rule2}
   !row=2
   for %series2 {%towers2}
      'Cannot have zeros for any multiplicative analysis, so recode 0 to 0.01
      {%series1}_{%series2}_n = @recode({%series1}_{%series2}_n<=0,0.01,{%series1}_{%series2}_n)
      
      smpl @first %history_end_date
         freeze(mode=overwrite,{%series1}_{%series2}_stats) {%series1}_{%series2}_n.ets(e=e, t=e, s=e,modsel=aic,graph=cl,table=cl) forecast
         
      smpl %forecast_start_date %forecast_end_date
         genr {%series1}_{%series2}_ets_f = @round(forecast)
         
         'Under circumstances of small numbers of observastions (i.e., medium wtc overflights), the ETS framework can make negative forecasts, so recode for this. This is very rare, just check for it
         {%series1}_{%series2}_ets_f  = @recode({%series1}_{%series2}_ets_f <0,0,{%series1}_{%series2}_ets_f )

         'Make an actual series that is only for the forecast period
         genr {%series1}_{%series2}_ets_a = {%series1}_{%series2}_n
         delete forecast
      smpl @all
      
      scalar {%series1}_{%series2}_ets_mape = @mape({%series1}_{%series2}_ets_f,{%series1}_{%series2}_ets_a)
      scalar {%series1}_{%series2}_ets_mae = @mae({%series1}_{%series2}_ets_f,{%series1}_{%series2}_ets_a)

      if %series1="IFR" then      
         a_ets_feval_ifr(!row,1) = %series2
         a_ets_feval_ifr(!row,2) = {%series1}_{%series2}_ets_mape
         a_ets_feval_ifr(!row,3) = {%series1}_{%series2}_ets_mae
         !row=!row+1
      else
         a_ets_feval_vfr(!row,1) = %series2
         a_ets_feval_vfr(!row,2) = {%series1}_{%series2}_ets_mape
         a_ets_feval_vfr(!row,3) = {%series1}_{%series2}_ets_mae
         !row=!row+1
      endif
   next
next
smpl @all
stop

Any idea what this error means: "Eviews 10 Has Stopped Working" Any way to find out what is really going on?

Note the table has been correctly made in commands prior to what you see here

-Bill

Posted: **Mon Jul 09, 2018 9:24 am**

Could you provide the workfile so we can run the code?

Posted: **Mon Jul 09, 2018 9:33 am**

Here is the data in a workfile. I had to make this separately since the program would crash.

Posted: **Mon Jul 09, 2018 9:45 am**

To run the code we'll need some of the program variables defined. %history_end_date, %forecast_start_date, etc...

Posted: **Mon Jul 09, 2018 10:31 am**

The program is attached

thanks...it was nice to see EViews in Boulder a few weeks back

Posted: **Mon Jul 09, 2018 11:06 am**

Was great being in Boulder. We love that conference.

The issue you're facing is a bug in EViews - we'll get it fixed. For what it is worth, it is during ETS smoothing of IFR_CYHC. I'll provide more details as we find them.

Posted: **Mon Jul 09, 2018 11:17 am**

I think it could be more of a general problem, since it still occurs if I drop CYHC. In the bottom left of the screen it is at the Optimizing Model stage.

Hate to ask, but will the bug fix be somewhat quick as we need to get some forecasts done. See the screenshot attached.

Cheers!

Posted: **Mon Jul 09, 2018 11:46 am**

The bug is that some ETS models do not converge, and then the likelihood value is NA, and the graphing code doesn't like graphing NAs. You can make a temporary fix by changing:

Code: Select all

freeze(mode=overwrite,{%series1}_{%series2}_stats) {%series1}_{%series2}_n.ets(e=e,t=e,s=e,modsel=aic,graph=cl,table=cl) forecast

to be:

Code: Select all

freeze(mode=overwrite,{%series1}_{%series2}_stats) {%series1}_{%series2}_n.ets(e=e,t=e,s=e,modsel=aic,graph=cl,table=cl, noconv) forecast

Or, getting rid of the graph=cl option.

Posted: **Mon Jul 09, 2018 12:24 pm**

thanks....what is the noconv option?

I did not see it in the online manual:

http://www.eviews.com/help/helpintro.ht ... d-ets.html

cheers!!!

Posted: **Mon Jul 09, 2018 1:08 pm**

Reject non-converged models.

EViews.com

Eviews 10 Has Stopped Working

Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working

Re: Eviews 10 Has Stopped Working