MCGRAPHIRF (revised)

[TomDoan]

This is a revised version of the MCGRAPHIRF procedure. MCGRAPHIRF organizes graphs for the output from various procedures which draw impulse response functions. Among the version 8 examples which use this are:

1. BOOTVAR.RPF (boot strapping a VAR)
2. MONTESUR.RPF (Gibbs sampling for a near-VAR)
3. MONTESVAR.RPF (Gibbs sampling for an overidentified SVAR)
4. MONTEVAR.RPF (standard Monte Carlo integration for an OLS VAR)

By separating the tasks of generating the impulse responses (done by a variety of methods) and organizing the graphs (done by MCGRAPHIRF), we can provide greater flexibility in the graphics. The newest option is

[COMMONSCALE]/NOCOMMONSCALE

which allows you (with NOCOMMONSCALE) to override the standard behavior of making all responses of a particular variable graphed using the same scale.

mcgraphirf.src

Procedure file

(11.98 KiB) Downloaded 160 times

If you want the calculations of the confidence bands without the graphs, use MCPROCESSIRF

http://www.estima.com/forum/viewtopic.php?f=7&t=996

[TWG]

Tom, I have one question, how can I get a table with the values of
the graph of the irf using mcgraphirf, with upper, lower and the center values.

Best regards


W

[TomDoan]

You would use MCPROCESSIRF. I edited the original message to include a link to it.

[WALLE]

Hi,

I'm running a 6 variable VAR as in below and I'm trying to compute the confidence intervals for Impulse responses. I looked up at the User guide and read up about the Monte Carlo procedure. I carried out the procedure using the following code and I obtained the graphs. However, the graphs are too many; 36 on a page resulting in the output being very tiny to read. And I really would like to have just the response of all variables to one shock per page instead of the present 6X6 format. How do I go about this?

Here is my code:

Code: Select all

***************************************************************
*SET UP/ESTIMATE VAR SYSTEM
***************************************************************
compute neqn = 6
compute nlags = 2
compute nsteps = 12
compute ndraws = 10000

system(model=canvar)
variables OILP CAN_CPI_INF CAN_GDEF_INF CAN_WAG_INF CAN_OUTPUT CAN_UR
lags 1 to nlags
det constant
end(system)

estimate(noprint,resids=varresidscan)

******************************************************************
*GENERATE IRF AND CONFIDENCE INTERVAL USING MONTE CARLO
******************************************************************
compute nvar   =%nvar
compute fxx    =%decomp(%xx)
compute fwish  =%decomp(inv(%nobs*%sigma))
compute wishdof=%nobs-%nreg
compute betaols=%modelgetcoeffs(canvar)

declare vect[rect] %%responses(ndraws)
declare rect[series] impulses(nvar,nvar)

infobox(action=define,progress,lower=1,upper=ndraws) "Monte Carlo Integration"
do draw=1,ndraws

   if %clock(draw,2)==1 {
      compute sigmad  =%ranwisharti(fwish,wishdof)
      compute fsigma  =%decomp(sigmad)
      compute betau   =%ranmvkron(fsigma,fxx)
      compute betadraw=betaols+betau
   }
   else
      compute betadraw=betaols-betau
   *
   * Push the draw for the coefficient back into the model.
   *
   compute %modelsetcoeffs(canvar,betadraw)
     results=impulses,steps=nsteps)
   *
   * Save the impulse responses
   *
   dim %%responses(draw)(nvar*nvar,nsteps)
   ewise %%responses(draw)(i,j)=ix=%vec(%xt(impulses,j)),ix(i)
   infobox(current=draw)
end do draw
infobox(action=remove)

@MCGRAPHIRF(MODEL=CANVAR)

[TomDoan]

Hi,

I'm running a 6 variable VAR as in below and I'm trying to compute the confidence intervals for Impulse responses. I looked up at the User guide and read up about the Monte Carlo procedure. I carried out the procedure using the following code and I obtained the graphs. However, the graphs are too many; 36 on a page resulting in the output being very tiny to read. And I really would like to have just the response of all variables to one shock per page instead of the present 6X6 format. How do I go about this?

Here is my code:

Code: Select all

***************************************************************
*SET UP/ESTIMATE VAR SYSTEM
***************************************************************
compute neqn = 6
compute nlags = 2
compute nsteps = 12
compute ndraws = 10000

system(model=canvar)
variables OILP CAN_CPI_INF CAN_GDEF_INF CAN_WAG_INF CAN_OUTPUT CAN_UR
lags 1 to nlags
det constant
end(system)

estimate(noprint,resids=varresidscan)

******************************************************************
*GENERATE IRF AND CONFIDENCE INTERVAL USING MONTE CARLO
******************************************************************
compute nvar   =%nvar
compute fxx    =%decomp(%xx)
compute fwish  =%decomp(inv(%nobs*%sigma))
compute wishdof=%nobs-%nreg
compute betaols=%modelgetcoeffs(canvar)

declare vect[rect] %%responses(ndraws)
declare rect[series] impulses(nvar,nvar)

infobox(action=define,progress,lower=1,upper=ndraws) "Monte Carlo Integration"
do draw=1,ndraws

   if %clock(draw,2)==1 {
      compute sigmad  =%ranwisharti(fwish,wishdof)
      compute fsigma  =%decomp(sigmad)
      compute betau   =%ranmvkron(fsigma,fxx)
      compute betadraw=betaols+betau
   }
   else
      compute betadraw=betaols-betau
   *
   * Push the draw for the coefficient back into the model.
   *
   compute %modelsetcoeffs(canvar,betadraw)
     results=impulses,steps=nsteps)
   *
   * Save the impulse responses
   *
   dim %%responses(draw)(nvar*nvar,nsteps)
   ewise %%responses(draw)(i,j)=ix=%vec(%xt(impulses,j)),ix(i)
   infobox(current=draw)
end do draw
infobox(action=remove)

@MCGRAPHIRF(MODEL=CANVAR)

@MCGRAPHIRF has quite a few options for formatting graphs. You might want to read the comments at the top of the MCGRAPHIRF.SRC file. What you're looking for is the option PAGE=BYSHOCK.

[WALLE]

Thanks Tom. Worked like magic