Clear Filters
Clear Filters

Total Least squares error help. I tried a couple different things here, but i'm not exactly sure how to do a total least squares error. The numbers aren't making sense to me.

3 views (last 30 days)
clear all
% A represents the sample measured in days
A = [1, 0; 1, 1; 1, 2; 1, 3; 1, 4; 1, 5; 1, 6; 1, 7]
A = 8×2
1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7
% b represents the amount of radium measured
b = [100; 82.7; 68.3; 56.5; 46.7; 38.6; 31.9; 26.4]
b = 8×1
100.0000 82.7000 68.3000 56.5000 46.7000 38.6000 31.9000 26.4000
% Least Squares
c = lsqr(A,log(b))
lsqr converged at iteration 2 to a solution with relative residual 9.5e-05.
c = 2×1
4.6051 -0.1903
g = A.'*A
g = 2×2
8 28 28 140
rhs = A.'*log(b);
x = g\rhs
x = 2×1
4.6051 -0.1903
thalf = -x(2)^-1*log(2)
thalf = 3.6417
m0 = exp(x(1))
m0 = 99.9942
m = @(t) m0*exp(x(2)*t);
predictionvalues = m([0, 1, 2, 3, 4, 5, 6, 7]);
squareddifferences = (predictionvalues - b).^2;
totalerror = sum(squareddifferences)
totalerror = 1×8
1.0e+04 * 1.9858 1.0169 0.5788 0.4646 0.5396 0.7175 0.9436 1.1847
t = [0, 1, 2, 3, 4, 5, 6, 7];
plot(t, x(1) + x(2)*t, t, log(b),'o')
%%plot(t, abs(exp(y_approx(t)-b)))
xlabel("t")
ylabel("error")
  2 Comments
Rian Sullivan
Rian Sullivan on 17 Feb 2024
Is the total least squares function okay? I was just making sure I even did the total least squares error code right. The total error seemed really small, but maybe im misunderstanding what it actually means. totalerror =
1.0e+04 *
Is what pops up along with all the other columns and their errors. It makes sense I believe.. but I'm just struggling with thinking my number of error doesn't look right.

Sign in to comment.

Answers (1)

David Goodmanson
David Goodmanson on 17 Feb 2024
Edited: David Goodmanson on 17 Feb 2024
Hi Rian,
the problem is that b is a column vector, but predictionvalues is a row vector. Using
predictionvalues = m([0, 1, 2, 3, 4, 5, 6, 7])'; % column vector
squareddifferences = (predictionvalues - b).^2;
totalerror = sum(squareddifferences)
totalerror =
0.0033
although depending on circumstances, you may want to use sqrt(totalerror/8), the rms value, as the final result
  1 Comment
Rian Sullivan
Rian Sullivan on 17 Feb 2024
Oh wow! That makes complete sense. Can't believe I missed that. I think that total error makes a lot more sense to me now. Thank you so much!

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!