Impulse from exogenous regressor in VAR

[luching]

Hi Tom, I am trying to include a set of exogenous regressors in an otherwise standard VAR model. That I think can be done by by simply augmenting them along with the constant term. I am however trying to find a way to draw impulse responses from those exogenous regressors. Any hint will be of great help.

[TomDoan]

For an example, see MONTEEXOGVAR.RPF.

[zw83189]

Dear Tom,

I used the tried to compute a shock from exogenous variable accroding to "MONTEEXOGVAR.RPF". I have changed the data name, dates, variable names. I have also changed the lag length into 1 lag, and the number of keeper draws (ndraws) to be 50 instead of 10000. Also I have estimated the model using the following commands:

Code: Select all

system(model=varmodel)
variables logrbtc logrltc
lags 1
det constant volbtc volltc tranbtc tranltc lwikibtc lwikiltc lhashbtc lhashltc
end(system)

My problem is after running the following command, the progress box which shows "Monte Carlo Integration" freezes there and does not seem like it is progressing.
infoxbox(action=define,progress,lower=1,upper=ndraws) "Monte Carlo Integration"

Is there any reason the above progress box is not working? Or it needs to take long time to progress? Is it because i have included too many exogenous variables (each of these series has about 800 observations)?




I tried to copy all the codes from MONTEEXOGVAR.RPF and modified a bit to suit my needs. and run all the command by pressing "execute instructions" button on the top. Then it shows the following contents: (there is no impulse response graphs)

Code: Select all

VAR/System - Estimation by Least Squares
Daily(7) Data From 2013:07:18 To 2015:10:06
Usable Observations                       811

Dependent Variable LOGRBTC
Mean of Dependent Variable       0.0013016151
Std Error of Dependent Variable  0.0485064172
Standard Error of Estimate       0.0472982349
Sum of Squared Residuals         1.7896984171
Durbin-Watson Statistic                1.9985

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  LOGRBTC{1}                         0.0200       0.0498      0.40124  0.68835094
2.  LOGRLTC{1}                        -0.0771       0.0307     -2.51244  0.01218574
3.  Constant                           0.0478       0.0354      1.35095  0.17709437
4.  VOLBTC                       -5.1453e-004  2.6010e-004     -1.97820  0.04824836
5.  VOLLTC                        1.4843e-004  1.9520e-004      0.76042  0.44722781
6.  TRANBTC                       1.5667e-006  4.2118e-007      3.71979  0.00021331
7.  TRANLTC                      -5.4267e-009  9.4356e-009     -0.57513  0.56536312
8.  LWIKIBTC                     -7.9859e-004  4.0385e-003     -0.19774  0.84329738
9.  LWIKILTC                     -9.1745e-003  4.5734e-003     -2.00605  0.04518741
10. LHASHBTC                      5.8074e-003  3.3216e-003      1.74838  0.08078261
11. LHASHLTC                          -0.0163  5.4967e-003     -2.96606  0.00310620

    F-Tests, Dependent Variable LOGRBTC
              Variable           F-Statistic     Signif
    *******************************************************
    LOGRBTC                            0.1610    0.6883509
    LOGRLTC                            6.3123    0.0121857


Dependent Variable LOGRLTC
Mean of Dependent Variable       0.0001472750
Std Error of Dependent Variable  0.0795439029
Standard Error of Estimate       0.0782936148
Sum of Squared Residuals         4.9039120937
Durbin-Watson Statistic                1.9866

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  LOGRBTC{1}                         0.1775       0.0824      2.15481  0.03147437
2.  LOGRLTC{1}                        -0.0696       0.0508     -1.37074  0.17084055
3.  Constant                      2.0541e-003       0.0586      0.03508  0.97202447
4.  VOLBTC                       -1.5734e-003  4.3055e-004     -3.65451  0.00027449
5.  VOLLTC                        1.3185e-003  3.2312e-004      4.08056  0.00004943
6.  TRANBTC                       2.6311e-006  6.9718e-007      3.77390  0.00017259
7.  TRANLTC                      -4.0716e-008  1.5619e-008     -2.60682  0.00930892
8.  LWIKIBTC                     -4.1473e-003  6.6851e-003     -0.62038  0.53518252
9.  LWIKILTC                     -4.3368e-003  7.5705e-003     -0.57286  0.56690296
10. LHASHBTC                      8.2068e-003  5.4983e-003      1.49261  0.13593438
11. LHASHLTC                          -0.0172  9.0988e-003     -1.89388  0.05860203

    F-Tests, Dependent Variable LOGRLTC
              Variable           F-Statistic     Signif
    *******************************************************
    LOGRBTC                            4.6432    0.0314744
    LOGRLTC                            1.8789    0.1708405

## OP3. This Instruction Does Not Have An Option FLA
>>>> flatten=<<<<

[TomDoan]

The problem is that the version of the program that you're using doesn't have the FLATTEN option on IMPULSE, which was added with version 9. You should talk to your department about getting their software updated.

[jabostool]

Hi.
I need help. I need plot the FEVD and the historical decomposition of the expgenous variables in a SVAR-X.

[TomDoan]

That makes no sense. Those only apply to the endogenous variables.

[luching]

Hi Tom, The model runs fine and I was able to get the impulse responses due to the exogenous variables. But I am a bit stuck with the historical decompositions. How should I modify the standard historical decomposition line used, for instance in montevar. Any hint will be great. Below is what I usually have for a montevar.

history(model=varmodel,noadd,results=history,cv=sigmad,from=hstart,to=hend)

[TomDoan]

You don't. The historical decomposition only applies to a self-contained model.