Documentation of searchbj

Global Index (short | long) | Local contents | Local Index (short | long)

Function Synopsis

[t1,e,el,vlc,w,V1,c,d,h,st]=searchbj(z,t,g,lim,V,nb,nc,nd,nf,nk,ni,display)

Help text

SEARCHBJ searches for lower values of the prediction error criterion

   [T1,E,EL,V,W,V1,C,D,H,ST] = searchbj(Z,T,G,LIM,V,NB,NC,ND,NF,NK)

   The routine searches for a lower value than V of the prediction error
   criterion for the Box-Jenkins model, starting at T, looking in the
   G-direction. T1 is returned as the parameters that give a lower value V1
   If no lower value is found, ST=1. C, D and H (=C*F) are the poly-
   nomials associated with T1, and E, EL, V and W are filtered data
   sequences. The routine is to be used as a subroutine to BJ. See
   BJ for an explanation of the other arguments.

Cross-Reference Information

Listing of function searchbj

function [t1,e,el,vlc,w,V1,c,d,h,st]=searchbj(z,t,g,lim,V,nb,nc,nd,nf,nk,ni,display)

%   L. Ljung 10-1-86, 9-25-93
%   Copyright (c) 1986-98 by The MathWorks, Inc.
%   $Revision: 2.3 $  $Date: 1997/12/02 03:44:03 $

l=0;,k=1;V1=V+1; n=nb+nc+nd+nf; st=0;
[mz,nz]=size(z(:,1));
ll=ones(mz,nz)*lim;
while [V1 > V l<10],
t1=t+k*g; if l==9,t1=t;end
      c=fstab([1 t1(nb+1:nb+nc).']);t1(nb+1:nb+nc)=c(2:nc+1).';
      f=fstab([1 t1(nb+nc+nd+1:n).']);t1(nb+nc+nd+1:n)=f(2:nf+1).';
      h=conv(c,f);
      d=[1 t1(nb+nc+1:nb+nc+nd).'];
      b=[zeros(1,nk) t1(1:nb).'];
     w=filter(b,f,z(:,2)); vlc=z(:,1)-w; e=pefilt(d,c,vlc,zeros(ni,1));
     if lim==0,el=e;else la=abs(e)+eps*ll;el=e.*(min(la,ll)./la);end
     V1=real(e'*el/(length(e)-ni));

if display,
   %home
   disp(int2str(l))
end
k=k/2;
l=l+1; if l==10,st=1;end
end