24 lines
643 B
Mathematica
24 lines
643 B
Mathematica
|
function [Omega] = getErrorFullCovariance(W, Y_tau, nbDOF)
|
||
|
|
|
||
|
|
% Authors: Quentin Leboutet, Julien Roux, Alexandre Janot and Gordon Cheng
|
||
|
|
%
|
||
|
|
% Gives the error covariance from the observation matrix following the
|
||
|
|
% approach of Han et al. (2020). This is used for IRLS computation.
|
||
|
|
|
||
|
|
[neq,np] = size(W);
|
||
|
|
|
||
|
|
Residual = zeros(nbDOF,neq/nbDOF);
|
||
|
|
|
||
|
|
for i=1:nbDOF
|
||
|
|
% Calculate OLS solution:
|
||
|
|
Beta_LS = W(i:nbDOF:end,:)\Y_tau(i:nbDOF:end);
|
||
|
|
|
||
|
|
% Compute error covariance:
|
||
|
|
Residual(i,:) = (Y_tau(i:nbDOF:end) - W(i:nbDOF:end,:)*Beta_LS).';
|
||
|
|
|
||
|
|
end
|
||
|
|
|
||
|
|
Omega = (Residual*Residual.')/(neq-np); % Compute the resulting error covariance estimate
|
||
|
|
|
||
|
|
end
|