function  [AR,RC,PE] = durlev(ACF);
 function  [MX,PE] = durlev(ACF);
 estimates AR(p) model parameter by solving the
 Yule-Walker with the Durbin-Levinson recursion
 for multiple channels
  INPUT:
 ACF	Autocorrelation function from lag=[0:p]

  OUTPUT
 AR    autoregressive model parameter	
 RC    reflection coefficients (= -PARCOR coefficients)
 PE    remaining error variance
 MX    transformation matrix between ARP and RC (Attention: needs O(p^2) memory)
        AR(:,K) = MX(:,K*(K-1)/2+(1:K));
        RC(:,K) = MX(:,(1:K).*(2:K+1)/2);

 All input and output parameters are organized in rows, one row 
 corresponds to the parameters of one channel

 see also ACOVF ACORF AR2RC RC2AR LATTICE
 
 REFERENCES:
  Levinson N. (1947) "The Wiener RMS(root-mean-square) error criterion in filter design and prediction." J. Math. Phys., 25, pp.261-278.
  Durbin J. (1960) "The fitting of time series models." Rev. Int. Stat. Inst. vol 28., pp 233-244.
  P.J. Brockwell and R. A. Davis "Time Series: Theory and Methods", 2nd ed. Springer, 1991.
  S. Haykin "Adaptive Filter Theory" 3rd ed. Prentice Hall, 1996.
  M.B. Priestley "Spectral Analysis and Time Series" Academic Press, 1981. 
  W.S. Wei "Time Series Analysis" Addison Wesley, 1990.

Package: tsa