When forcing intercept to zero, why JMP does not provide R-squared?

JK — Fri, 09 Jun 2023 00:24:05 GMT

Hello, I have some questions.

When I have a linear model in JMP, I want to force intercept to zero.

So, In Fit Special, I chose "Constrain Intercept to 0" then I got the new output from JMP.

My questions are

1) why JMP does not provide R-squared when intercept becomes zero?

2) R-squared is calculated by SSR/SST, can I calculate R-squared by myself using the data JMP provides? If I calculated by myself, it's higher than before. Is this possible? In this case, it says R-squared is 99%. Can I say this R-squared is correct?

Thanks,

Re: When forcing intercept to zero, why JMP does not provide R-squared?

Mark_Bailey — Wed, 04 Nov 2020 13:44:59 GMT

This case is one of "just because you can doesn't mean you should."

Yes, the sum of squares that are normally used to compute R square are provided. But the resulting ratio has no meaning. R square is called the "coefficient of determination" in linear regression. When R square equals 1, then all the variation in Y is explained or determined by variation in X. (That is not a statement about causation.) When R square equals 0, then Y is independent of X. It is, therefore, the mean of Y for all X. But your null hypothesis is that Y = 0. So the ANOVA might be statistically significant, but the R square is undefined.

See this article for more information.

topic When forcing intercept to zero, why JMP does not provide R-squared? in Discussions

When forcing intercept to zero, why JMP does not provide R-squared?

Re: When forcing intercept to zero, why JMP does not provide R-squared?