matlab LSQ error calculation computation time

Hi everybody,

I am trying to calculate a simple least squares solution and its corresponding error with the function below. When calculating the error the script slows down significantly. In fact a run with an active profiler shows that 99% of computation time is spent in the line

err=psi'*(I-H*t)*psi.

Can anybody explain why this line is so expensive or give an solution that is faster?

%H: observation matrix,
%I: identiy matrix,
%psi: measurement vector,
%PCO: parameter estimate,
%err: error of the estimate
function [PCO,err]=calcPC_LSQ(H,I,psi)
    t=(H'*H)\H';
    PCO=t*psi;
    err=psi'*(I-H*t)*psi;
end