g13bdf
g13bdf
© Numerical Algorithms Group, 2002.
Purpose
G13BDF Multivariate time series, preliminary estimation of transfer
function model
Synopsis
[wds,isf,ifail] = g13bdf(r0,r,nna,s<,ifail>)
Description
The routine calculates estimates of parameters
(delta) ,(delta) ,...,(delta) , (omega) ,(omega) ,...,(omega) in
1 2 p 0 1 q
the transfer function model
y =(delta) y +(delta) y +...+(delta) y +(omega) x
t 1 t-1 2 t-2 p t-p 0 t-b
-(omega) x -...-(omega) x
1 t-b-1 q t-b-q
given cross correlations between the series x and lagged values
t
of y :
t
r (l) , l=0,1,...,L
xy
and the ratio of standard deviations s /s , as supplied by G13BCF.
y x
It is assumed that the series x used to calculate the cross
t
correlations is a sample from a time series with true
autocorrelations of zero. Otherwise the cross correlations
between the series b and a , as defined in the description of
t t
G13BAF, should be used in place of those between y and x .
t t
The estimates are obtained by solving for
(delta) ,(delta) ,...,(delta) the equations
1 2 p
r (b+q+j)=(delta) r (b+q+j-1)+...+(delta) r (b+q+j-p) ,
xy 1 xy p xy
j=1,2,...,p
then calculating
(omega) =+-(s /s ){r (b+i)-(delta) r (b+i-1)-...
i y x xy 1 xy
-(delta) r (b+i-p)} , i=0,1,...,q
p xy
where the '+' is used for (omega) and '-' for (omega) , i>0.
0 i
Any value of r (l) arising in these equations for l<b is taken
xy
as zero. The parameters (delta) ,(delta) ,...,(delta) are
1 2 p
checked as to whether they satisfy the stability criterion.
Parameters
g13bdf
Required Input Arguments:
r0 real
r (:) real
nna (3) integer
s real
Optional Input Arguments: <Default>
ifail integer -1
Output Arguments:
wds (:) real
isf (2) integer
ifail integer