Page 1 of 1

BVARs and fcast

Posted: Tue Aug 16, 2016 6:54 am
by Sharky
Hi

I am using the fcast add-on to forecast with VARs. It works fine with the ECM (.ec) but when I tell it to fcast all the series using the .bvar, several of the series don't forecast out of sample. Oddly, one (the first) does.

Does anyone have any idea whether there are any known issues in using fcast with BVARs?

Re: BVARs and fcast

Posted: Tue Aug 16, 2016 8:50 am
by Sharky
I've tried using makemodel and solving for forecasts that way but get the same problem so maybe this shouldn't be in "Add-in support"...

Code: Select all

for !ii=!m to !b-1 'EXPAND

smpl @first @first+!ii
var brecvar.bvar(NOCONST,prior=lit) 1 4 c_ inve y labobs     robs    pinfobs w
brecvar.makemodel(bvarmodel)


smpl @first+1+!ii @last
bvarmodel.solve
brecvar.fcast _bf
brecvartab(!ii-!m+1,1) = c__0(!ii+2)
brecvartab(!ii-!m+1,2) = inve_0(!ii+2)
brecvartab(!ii-!m+1,3) = y_0(!ii+2)
brecvartab(!ii-!m+1,4) = labobs_0(!ii+2)

delete c__bf inve_bf y_bf labobs_bf robs_bf pinfobs_bf w_bf
next


gives zeros for the forecast period for inve_0

Code: Select all

Last updated: 08/16/16 - 16:46
Modified: 1988Q4 2014Q3 // bvarmodel.solve

0.000000
1.745513
3.931802
6.590199
8.559823
7.236351
5.592367
2.197469
-0.715824
0.886564
0.644268
2.695389
4.685949
3.611336
3.935830
5.879947
8.245010
7.782273
7.867575
4.651494
2.420262
0.073466
-0.063016
-1.603982
-0.026416
2.730538
3.720697
5.191887
7.284921
8.344859
8.523570
12.20743
15.28286
15.30807
14.20564
11.52735
8.316766
6.661362
4.632351
-1.573486
-7.648966
-9.031301
-8.184192
-7.525205
-4.495301
-3.245115
-3.032536
0.394131
3.603058
8.345976
9.941303
10.62400
11.00101
17.25353
19.17401
20.23301
20.25606
18.62860
19.66837
17.83569
15.80006
5.647101
5.482860
8.201159
8.303956
8.430868
7.549430
5.756958
1.970574
-2.368052
-6.878626
-8.601699
-8.442075
-6.353533
-2.605177
1.615638
2.955661
5.846936
6.564426
7.002119
6.383877
6.279089
5.056519
5.948848
5.245713
4.639691
3.900586
3.858780
0.713716
1.097559
1.655001
0.763065
0.053990
0.907241
-0.093625
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000



When I run the same code with .ec instead of .bvar it works fine.

Re: BVARs and fcast

Posted: Tue Aug 16, 2016 9:11 am
by EViews Gareth
Which version/build date of EViews?

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 3:23 am
by Sharky
Eviews 8 (x64)

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 8:02 am
by EViews Gareth
Build date? (Help->About EViews).

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 8:11 am
by Sharky
Mar 22 2013

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 8:36 am
by EViews Gareth
That's very out of date. Update to the latest version.

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 8:40 am
by Sharky
I've got it through my university. Is it possible to update without them doing so for the whole department? Or do I need to wait for that?

Re: BVARs and fcast

Posted: Wed Aug 17, 2016 8:42 am
by EViews Gareth
I don't know the details of your particular install. Try it and see?