Regression line that includes x and y errors.

Hi,

I have x and y data set, and each x and y variables have 5% and 10% uncertainty or error, respectively. How I should find a one regresion or best fit line (Best_y = mx + c) for such a dataset? Here is the dataset. Any thoughts? Thanks.

Answer 1

[Matt J]

I would recommend linear2dFit() form this FEX download,

https://www.mathworks.com/matlabcentral/fileexchange/87584-object-oriented-tools-for-fitting-conics-and-quadrics

fitResult=linear2dFit([x,y]')

[Research]

Thanks. I tried to run the example code, but got error on this line: xy=linear2dFit.xysim(p1, p2, t, sig). The error is The specified superclass 'conicFit' contains a parse error, cannot be found on MATLAB's search path, or is shadowed by another file with the same name.

I found that I dont have "linear2dFit.xysim" FEX function. I'm missing anything. Thanks.

[Matt J]

It sounds to me like you only downloaded linear2dFit. You need to download all the files.

[Research]

Thanks. It runs now. This is a great function. So, just I understand the linear2dFit function, particularly how the function incorporates the error values. In my case, I need to build an array (xy here) that includes +/-x and +/-y error values.This xy array has x and y values also error dx and dy values too. I believe the function then find a best fit through all these points. Thanks!

[Matt J]

Thanks. It runs now. This is a great function

Glad it's working for you, but if it's doing what you need, I'd be much obliged if you'd Accept-click the answer.

particularly how the function incorporates the error values?

The fit is a total least squares fit, so it assumes there are errors in both x and y.