Page 1 of 1

Different results from different graph commands

Posted: Sat Apr 23, 2016 3:26 am
by indrani_5
I am working on a basic recursive VAR with standard specification. When I generate impulse responses with the command 'montevar', I get a different set of impulse responses than when I generate impulse responses using 'mcgraphirf'. For example, using the specification - npa lp y d p comm_e int_rate sp , I find that responses which were significant under 'montevar' became insignificant under'mcgraphirf''. Secondly,since this is recursive VAR, npa should not contemporaneously respond to lp shock but other variables which follow lp should respond contemporaneously to lp shock. But I find that these variables impact response is zero when i use 'montevar'. How do I make my results consistent? The graphs and data are attached. The code is given below

Code: Select all

[/OPEN DATA "E:\phd\research\RATS\bubble_basic.xls"
CALENDAR(Q) 1984:4
DATA(FORMAT=XLS,ORG=COLUMNS,SHEET="Sheet2") 1984:04 2008:02 y d p lp sp hp m2 comm_e comm_ne int_rate lp_nd $
 lp_nb nltl npa nltl_1 nltl_exp



system(model=varmodel)
variables  npa lp y d p comm_e int_rate sp
lags 1
det constant
end(system)
estimate
@VARLagSelect(lags=12, crit=aic)
# y d p comm_e nltl_1 lp int_rate sp
@VARLagSelect(lags=12, crit=sbc)
# y d p comm_e nltl_1 lp int_rate sp
*@VARIRF(model=varmodel,steps=15,page=byshocks)
*@montevar(draws=10000, model=varmodel)

source mcvardodrawsunit.src
@MCVARDoDraws(model=varmodel,draws=1000,steps=10)
@mcgraphirf(model=varmodel,center=median,percent=||.025,.975||,$
shocks=||"NPA","LP","GDP","Dividends","Inflation","Commod.prices","int_rate","Share Prices"||,$
footer="95% Monte Carlo bands",page=byshocks)]

Re: Different results from different graph commands

Posted: Sat Apr 23, 2016 8:27 am
by TomDoan
source mcvardodrawsunit.src

switches to unit shocks from Cholesky shocks, so it will give completely different results. Since you're not interested in unit shocks, get rid of that.

Where did you find the use of mcvardodrawsunit.src?

Re: Different results from different graph commands

Posted: Sat Apr 23, 2016 12:42 pm
by indrani_5
Thank you for your answer. Could you please also comment on why other variables that follow the shocked variable do not respond contemporaneoulsy to the shock.


mcvardodrawsunit.src is in the varworkbook.

Re: Different results from different graph commands

Posted: Sat Apr 23, 2016 1:30 pm
by TomDoan
That's what unit shocks do. You're looking at the wrong example in the 1st edition of the VAR workbook. That's an intentionally simple model to demonstrate the different ways to do error bands.