topic Re: Can you display R squared value with Bi-variate fit special and forcing intercept through y = 0? in Discussions

Can you display R squared value with Bi-variate fit special and forcing intercept through y = 0?

Andrew_Botfield — Sun, 11 Jun 2023 12:04:05 GMT

Re: Can you display R squared value with Bi-variate fit special and forcing intercept through y = 0?

Byron_JMP — Tue, 22 Sep 2020 18:36:16 GMT

Yes, but No. Use the Fit Special option

(earlier I added the wrong screen shot)

The thing is, when you fix the slope or intercept, then you're not really doing a least squared fit.

Since RSquare is the sum of squares for the model divided by the total sum of squares, there isn't exactly a model to work with for the fit.

Re: Can you display R squared value with Bi-variate fit special and forcing intercept through y = 0?

Dan_Obermiller — Tue, 22 Sep 2020 23:58:27 GMT

Caution! Care must be taken when interpreting an RSquare value with the intercept forced to be 0. The usual statistics like RSquare assume that you are using the mean of the response (Y-bar) for your sums of squares. By forcing the intercept to be 0, you are replacing Y-bar with 0. This alters the typical interpretations of those statistics. How much is altered will depend on how far your Y-bar is from 0.

For example, it is possible to achieve a negative RSquare value when you constrain the intercept to 0. For this reason, most people are in favor of not reporting RSquare in such cases.