Bootstrapping and graphing VAR confidence intervals

[ylijohtaja]

Hi

My aim is to bootstrap (using Eviews 7) and graph impulse responses and confidence bands from a VAR model. Since I would like to use GIRFs instead of SVAR, I have modified the program bqboot.prg to correspond for GIRFs. The modified program appears to work well (would be great if someone could check if the modification yields the desired results, though) up to a three variable VAR. However, when a fourth variable is included in the system, the following error message emerges: "unable to compute due to missing data in "y= @coeff(1)*x1(-1) + @coeff(2)*x2(-1) + @coeff(3)*x3(-1) + @coeff(4)*x4(-1) + @coeff(5)" in "DO_MOD1.SOLVE"." I have been unable to figure out the reason for the error. (bootstrap4.prg)

Moreover, when trying to create the graphs regarding a two or three variable VAR the following error message appears: "SUB_BOOTIRF is not a defined subroutine in "CALL SUB_BOOTIRF(RSP_LVL,LVL_UPP,LVL_LOW,"FIG_LVL")". What could be the problem? (bootstrapbands.prg)

Thank you in advance!

[EViews Gareth]

Both of your programs use the CALL command to call a subroutine (call sub_bootirf). However nowhere have you defined what that subroutine is.

[spikestrang]

Hi,

I am trying to do a similar thing in graphing bootstrapped confidence intervals for a SVAR. I have edited a program found on the internet to suit my needs and but I get the error message ' FIG_LVL is not defined in "DO_FIG_LVL.DRAW(LINE,LEFT,RGB(180,180,180)) 0"' and cannot work out how to correct it. I have attached the program I am trying to run. Any help on this would be great.

Thanks.