Calculate adjusted R2 for constant model

Is there a way of calculating the adjusted R2 of a constant model (y= constant) for a set of fitted data in Matlab?

Answer 1

[Matt J]

R2=1-mean((y-constant).^2)/var(y);

[Haojue Yu]

Thanks! And what about adjusted R2? (So there is no table for matlab to show automatically? Like in the fitlm, it can do that.)

[Matt J]

And what about adjusted R2?

I believe for a constant model, R2 and adjusted R2 are the same.

(So there is no table for matlab to show automatically? Like in the fitlm, it can do that.)

With a constant model, fitlm will just estimate the constant as the mean of the data, giving R2 and adjusted R2 equal to zero. Since R2 is trivial in this case, I guess the ouput display doesn't show it, but you can still access it from the output, e.g.,

mdl=fitlm(1:10,rand(1,10),'constant');
mdl.Rsquared

ans = struct with fields:
    Ordinary: 0
    Adjusted: 0

[Haojue Yu]

Thank you very much!