correlation coefficient for a non-linear fit

Does anyone know how to compute the correlation coefficient between two vectors, which are not linearly correlated? At the moment, I use a polynomial fit (order 3) between two vectors. It seems that corrcoeff compute it only for linear fits. Thanks a lot!

Answer 1

[the cyclist]

There are several options for correlation coefficients besides Pearson's (which is probably the most commonly used). Did you want to compute Kendall's tau or Spearman's rho? If so, then you can use the function CORR, and set the appropriate parameter. If that's not what you want, you'll need to supply some more details. (You might try searching the File Exchange, as well.)

Answer 2

[Richard Willey]

Hi Marina

You might find the following blog posting of interest. (It shows some interesting ways to use Curve Fitting Toolbox)

http://blogs.mathworks.com/loren/2011/01/13/data-driven-fitting/

Answer 3

[Mariana]

Thanks! It seems (sorry for forgetting some statistics) that correlation coefficient is defined as a strength between two variables which are linearly correlated. I would like to have something describing the "goodness of a fit" (which can be linear, polynomial, exponential etc). I will take a look at the curve fitting toolbox.

Answer 4

[Mariana]

I found. "fit" does the job (we get rsquare among the statistics).

[Ameen Bassam]

Is rsquare valid for nonlinear regression ??

It is invalid according to this

https://statisticsbyjim.com/regression/r-squared-invalid-nonlinear-regression/