I think there are errors in drawing sign-restricted IRFs when the number of exogenous variables is more than two.

[naffe15]

I think there are errors in drawing sign-restricted IRFs when the number of exogenous variables is more than two.

[naffe15]

Hi,
the codes seems to work fine when computing IRFs with sign and narrative restrictions and with exogenous variables. This was not the case for the IRF with zero-sign restrictions. I adjusted the codes so that one can compute IRF with sign-zero restrictions with exogenous variables.

[SangJL]

Hello,

If I changed codes in exmple_6_VARX.m to include two exogenous variables and to use sign restrictions as following, current bar_.m doesn't work well in producing IRFs

% exogenous variable
z = demean(100*diff(log ([cpi_us str_us] )));
% include one lag for the exogenous variable
options.controls = lagX(z,[0:1]);
options.signs{1} = 'y(4,1:3,4)>0';
options.signs{2} = 'y(2,1:3,4)<0';

% estimate the VARX
bvar1 = bvar_(y,lags,options);

indx_sho = 4;
indx_var = [1, 2, 3, 4];
irfs_to_plot = bvar1.irsign_draws(:,:,indx_sho,:);
%%%%

But I found that bar_.m works well with my suggestions.

[naffe15]

Hi,
thanks for the follow up. now I understand where the problem is. the functions calculating the IRFs were computing the wrong number of lags when more than one exogenous variable is used. This problem is fixed now. thanks for spotting this.
f.

[SangJL]

My pleasure!

[SangJL]

Could I ask one more improvement?

Even though I set "options.K = 1000;", I may not get 1,000 draws because d =0 in "iresponse_sign.m" for cases. In my experiments, I can see the warning,"I could not find a rotation satisfying the restrictions.", for about 5~10% draws out of total designated draws(options.K). So, could you improve your "bvar_.m" to ensure that I can get initially-designed number of draws(options.K).

[naffe15]

I will look into it.
f.

[SangJL]

Appreciated!