Small sample bias in VARs

[comac]

Dear Tom,

Following the post on “Bootstrapping VAR residuals” (http://www.estima.com/forum/viewtopic.php?f=4&t=669) I am trying to compute the small sample bias for my VAR coefficients. I am taking the RATS example from Lutkepohl's New Introduction to Multiple Time Series Analysis.

Last time you suggested me to compute the matrix total as the bootstrapped estimate of the bias. Unfortunately, the results change evrytime I run the code (which is not what I really want)

I have also tried to play around with the code below but nothing happened. Do you have some suggestion?

Code: Select all

open data e1.dat
calendar(q) 1960
data(format=prn,org=columns,skips=6) 1960:01 1982:04 invest income cons
*
set dinc  = log(income/income{1})
set dcons = log(cons/cons{1})
set dinv  = log(invest/invest{1})
*
system(model=varmodel)
variables dinv dinc dcons
lags 1 2
det constant
end(system)
estimate(sigma,resids=resids) * 1978:4
compute base=%modelgetcoeffs(varmodel)
compute total=%zeros(%rows(base),%cols(base))
*
@VARBootSetup(model=varmodel) bootvar
*
compute rstart=%regstart()
compute rend  =%regend()
*
compute bootdraws=2000
compute nvar  =3
compute nsteps=8
*
infobox(action=define,progress,lower=1,upper=bootdraws) $
   "Bootstrap Simulations"
do draw=1,bootdraws
   @VARBootDraw(model=varmodel,resids=resids) rstart rend
   *
   * Estimate the model with resampled data
   *
   estimate(noprint,noftests)
   compute total=total+(%modelgetcoeffs(bootvar)-base)
   infobox(current=draw)
end do draw
infobox(action=remove)
*
compute total=total/bootdraws

[/quote]

Many thanks

C.

[TomDoan]

I'm a bit confused about your question. Are you concerned that you get a different bootstrap estimate each time you run a bootstrap? If you think the variability is too high, run more draws. If you want to control the draws, use the SEED instruction.

[comac]

Tom -- sorry if I was unclear. Anyway, I wanted exactly to set the seeds for the random number generator. Thanks a lot.