Markov switching model
Posted: Fri Jul 18, 2014 11:17 am
Dear Tom,
Please see the code below; I get unassigned error at the end of the Maximize instruction, something is missing but I can't figure out what is, could be my choice of starting values probabilities, please advise
Thanks
Please see the code below; I get unassigned error at the end of the Maximize instruction, something is missing but I can't figure out what is, could be my choice of starting values probabilities, please advise
Thanks
Code: Select all
source "C:\ Desktop\ markov.src"
OPEN DATA "C: \ Desktop\ returns_100.txt"
ALL 23005
DATA(FORMAT=free,ORG=COL) 1 23005 x
compute n=2
dec rect p(n-1,n)
dec vect f(n) pstar(n)
nonlin(parmset=markovparms) p
nonlin(parmset=mixparms) mux muT sigmax1 sigmaT1 sigmaTx1 sigmax2 sigmaT2 sigmaTx2
frml markov = $
k1 = sqrt(sigmaT1^2-sigmax1^(-2)*sigmaTx1^2), $
k2 = sqrt(sigmaT2^2-sigmax2^(-2)*sigmaTx2^2), $
q1 = sqrt(sigmaT1^2+sigmax1^2-2*sigmaTx1), $
q2 = sqrt(sigmaT2^2+sigmax2^2-2*sigmaTx2), $
g1 = (1.0/sigmax1)*%density((x{0}-mux)/sigmax1), $
g11 = -(1-sigmax1^(-2)*sigmaTx1)*(x{0}-mux)/k1, $
g12 = -(muT-mux)/k1, $
m1 = 1.0/%cdf(-(muT-mux)/q1), $
g2 = m1*g1*%cdf(g11+g12),$
h2 = (1.0/sigmax2)*%density((x{0}-mux)/sigmax2), $
h21 = (1-sigmax2^(-2)*sigmaTx2)*(x{0}-mux)/k2, $
h22 = (muT-mux)/k2, $
m2 = 1.0/%cdf((muT-mux)/q2), $
h2 = m2*h2*%cdf(h21+h22), $
(f = || g2, h2 ||) ,$
pstar=%msupdate(f,%mcstate(p,pstar),fpt), log(fpt)
compute p = ||.70,.80||, mux=2.792523, muT =0.0, sigmax1 =107.042328, sigmaT1 =100.0 , sigmaTx1 =0.0, sigmax2 =100.0, sigmaT2 =100.0, sigmaTx2 =0.0
maximize(parmset=markovparms+mixparms, startup = %(pstar=%mcergodic(p)), pmethod=simplex, piters=5, iters = 5000, method=bfgs, cvcrit=0.000001) markov