Robust orthogonal or total least squares linear regression

I am searching for a robust alternative for an orthogonal linear regression (or total least squares regression). I want to fit a line to 2D data, plot it in a x-y-graph together with the original data and calculate the slope of the fitted line, to use it for further calculations. I managed to use either a standart robust linear regression (robistfit) or a standart orthogonal regression (OrthLLS2D or wols), but not a combination of both. I also tried to use PCA (princomp), but there I do not manage to plot a line nor calculate the slope.

Thanks for any help! Martina

Answer 1

[Richard Willey]

Unfortunately, Statistics Toolbox doesn't currently offer a robust version of PCA.

Your best option would probably be to use Optimization Toolbox and code up a custom objective function using an L1 norm.