R-squared analysis of scatter plot resulting in NaN

Hi there,

I have two variables (eP and T) each of which are 1x 1244 in size. I have created a scatter plot of eP against T and I have fit a line. However, both variables have some NaN values. When I do an r-squared test I get an R squared value of NaN. Does anyone know how to perform an r-squared test and ignore NaN values?

Answer 1

[dpb]

'Pends on how you did it -- show your code for exact answer to that direction, but --

If you have the Statistics TB, then use

mdl=fitlm(x,y,'linear');        % will ignore NaN automagically
rsqO=mdl.Rsquared.Ordinary;     % ordinary r-squared
rsqA=mdl.Rsquared.Adjusted;     % adjusted r-squared

will handle it transparently.

Otherwise,

isOK=all(isfinite([ep T]),2);     % indices of those with NaN in neither variable

use ep(isOK), T(isOK) in your other code.