crosscor.m

偏最小二乘算法在MATLAB中的实现

function crcor = crosscor(x,y,n,period,flag)
%CROSSCOR Cross correlation of time series
% Performs the crosscorrelation function of two time series.
% The inputs are the two time series (x), and (y) and the
% number of sample periods (n) to consider. The sample period
% (period) is an optional input variable used to scale the 
% output plot and (flag) in an optional input variable which 
% changes the routine from cross correlation to cross covariance 
% when set to 1.
% The I/O format is: crcor = crosscor(x,y,n,period,flag);

%  Copyright
%  Barry M. Wise
%  1991
%  Modified November 1993
[mp,np] = size(x);
[mpy,npy] = size(y);
if np > mp
  x = x';
  mp = np;
end
if npy > mpy
  y = y';
  mpy = npy;
end
if mpy ~= mp
  error('The x and y vectors must be the same length')
end
crcor = zeros(2*n+1,1);
if nargin < 5
  flag = 0;
end
if flag == 1
  ax = mncn(x);
  ay = mncn(y);
else
  ax = auto(x);
  ay = auto(y);
end
for i = 1:n
  ax1 = ay(1:mp-n-1+i,1);
  ax2 = ax(n+2-i:mp,1);
  crcor(i,1) = ax1'*ax2/(mp-n+i-2);
  ax1 = ax(1:mp-n-1+i,1);
  ax2 = ay(n+2-i:mp,1);
  crcor(2*n+2-i,1) = ax1'*ax2/(mp-n+i-2);
end
crcor(n+1,1) = ax'*ay/(mp-1);
if nargin == 3
  scl = -n:1:n;
else
  scl = period*(-n:1:n);
end
plot(scl,crcor)
f = axis;
hold on
plot([f(1) f(2)],[0 0],'--g',[0 0],[f(3) f(4)],'--g')
if flag == 1
  title('Crosscovariance Function')
  ylabel('Covariance [CCF(Tau)]')
else
  title('Crosscorrelation Function')
  ylabel('Correlation [CCF(Tau)]') 
end
xlabel('Signal Time Shift (Tau)')
hold off