return error covariance constraints for serially correlated data
FORMAT [C] = pr_spm_ce(v,a)
v - (1 x l) v(i) = number of observations for ith block
a - AR coefficient expansion point (default a = [])
C{1} = h(1)*AR(a)
C{2} = h(1)*AR(a) + h(2)*dAR(a)/da(1);
C{3} = h(1)*AR(a) + h(2)*dAR(a)/da(1) + h(3)*dAR(a)/da(2);
See also: spm_Q.m
___________________________________________________________________________
Copyright (C) 2005 Wellcome Department of Imaging Neuroscience
0001 function [C] = pr_spm_ce(v,a) 0002 % return error covariance constraints for serially correlated data 0003 % FORMAT [C] = pr_spm_ce(v,a) 0004 % v - (1 x l) v(i) = number of observations for ith block 0005 % a - AR coefficient expansion point (default a = []) 0006 % 0007 % C{1} = h(1)*AR(a) 0008 % C{2} = h(1)*AR(a) + h(2)*dAR(a)/da(1); 0009 % C{3} = h(1)*AR(a) + h(2)*dAR(a)/da(1) + h(3)*dAR(a)/da(2); 0010 % 0011 % See also: spm_Q.m 0012 %___________________________________________________________________________ 0013 % Copyright (C) 2005 Wellcome Department of Imaging Neuroscience 0014 0015 % Karl Friston 0016 % $Id: spm_Ce.m 190 2005-06-20 07:04:35Z klaas $ 0017 0018 0019 0020 % defaults 0021 %--------------------------------------------------------------------------- 0022 if nargin == 1 0023 a = []; 0024 end 0025 0026 0027 % create blocks 0028 %--------------------------------------------------------------------------- 0029 C = {}; 0030 l = length(v); 0031 n = sum(v); 0032 k = 0; 0033 if l > 1 0034 for i = 1:l 0035 dCda = pr_spm_ce(v(i),a); 0036 for j = 1:length(dCda) 0037 [x y q] = find(dCda{j}); 0038 x = x + k; 0039 y = y + k; 0040 C{end + 1} = sparse(x,y,q,n,n); 0041 end 0042 k = v(i) + k; 0043 end 0044 else 0045 % dCda 0046 %================================================================== 0047 C{1} = pr_spm_q(a,v); 0048 dCda = pr_spm_diff('pr_spm_q',a,v,1); 0049 for i = 1:length(a) 0050 try 0051 C{i + 1} = dCda{i}; 0052 catch 0053 C{i + 1} = dCda; 0054 end 0055 end 0056 0057 end