armodelcov

Calculates the process autocovariance of an autoregressive model.

Description

Calculates the process autocovariance of an autoregressive model.
 
  USAGE:
  C=armodelcov(a)
 
  a: A row vector with AR coefficients.
     The notation follows that of other matlab functions such as arburg & filter.
     Note: this notation is different from what is commonly found in textbooks
           (sign is different and the length is one longer)
           x(t)=sum(-a(2:end).*x(t-(1:length(a)-1)))+randn;
  C: autocovariance matrix of same order as the ar model.
 
  Assumes that the noise term is IID gaussian noise (zero mean, unit stddev)
 
 
  EXAMPLE:
  ------------
  a=[1.0 -0.5 0.2 -0.0 -0.0 0.2 -0.5];
  n=length(a)-1;
  C_theory=armodelcov(a);
  C_theory=C_theory(:,1)'
  
  %generate noise according to the ar model and check what kind of autocov. it has.
  noise=filter(1,a,randn(10000,1));
  noise(1:100)=[]; %cut away edge effects.
  C_noise=xcov(noise,7,'unbiased')';
  C_noise(1:n)=[]

• MATLAB release: MATLAB 7 (R14)