Page 1 of 1

Restrictions on the conditional covariances

Posted: Sun Mar 30, 2025 11:26 am
by tx_ecnmst
I am trying to estimate a 3-variable VAR-GARCH-M with restrictions imposed on the conditional covariances using HADJUST. The problem is that I need to impose two different linear restrictions, varr=(hhs(1,1)+hhs(2,2)-2*hhs(2,1)) and covr=(hhs(3,2)-hhs(3,1)), and I cannot figure out the proper syntax to do so.

Code: Select all

set covr = 0.0
set varr = 0.0
dec symm[series] hhs(3,3)
clear(zeros) hhs
equation(REGRESSORS) infeq dpi 13
# d3{1 to 13} dc{1 to 13} realr3{1} hhs(1,1) hhs(3,1) varr covr
equation(REGRESSORS) nomeq d3 13
# dpi{1 to 13} dc{1 to 13} realr3{1} hhs(1,1) hhs(3,1) varr covr
equation(REGRESSORS) coneq dc 13
# d3{1 to 13} dpi{1 to 13} realr3{1} hhs(1,1) hhs(3,1) varr covr
group garchm infeq nomeq coneq

garch(model=garchm,rvectors=rd,hmatrices=hh,derives=dd,$
method=bfgs,pmethod=simplex,piters=500,mvhseries=hhs,$
mv=cc,stdresids=rstd,factorby=eigen,iters=1000,subiters=1000,$
hadjust=(varr=(hhs(1,1)+hhs(2,2)-2*hhs(2,1))(covr=(hhs(3,2)-hhs(3,1)))))

Re: Restrictions on the conditional covariances

Posted: Mon Mar 31, 2025 4:55 pm
by TomDoan
If you need more than one expression in an option like that, you enclose in %()---so

hadjust=%(varr=hhs(1,1)+hhs(2,2)-2*hhs(2,1),covr=hhs(3,2)-hhs(3,1))

Re: Restrictions on the conditional covariances

Posted: Mon Mar 31, 2025 11:30 pm
by tx_ecnmst
Thanks, again!