R^2 is negative ???

[seraphin]

I had the following pairs of numbers
(1, 0.11);(2, 0.138);(3, 0.164);(4, 0.176)
and asked EXCEL to plot them on scatter plot - x axis being 1,2,3,4 and y axis being 0.11, 0.138, 0.164, 0.176 respectively.

The original R^2 between X & Y for these pairs is 0.9724. However, if I set the intercept to be at (0,0), the R^2 is shown to be -1.167.
I thought R^2 cannot be negative. What's wrong with the calculation ? (I actually didn't calculate it as EXCEL 2010 can automatically produce these numbers)

Any help ??
Great many thanks

[newdoverman]

A sample workbook showing what you are actually doing would help. Can you post one?

[seraphin]

Not sure exactly how to do a workbook, but here is what I've tried (please click on the JPG file for the two different plots - one without setting intercept, one with setting the intercept at (0,0) ). You'd see that when the intercept is set at (0,0), R^2 = -1.167

1 0.11
2 0.138
3 0.164
4 0.176




Presentation1.jpg

[shg]

See http://www.excelforum.com/excel-gene...-to-0-0-a.html

[shg]

One of the links at the link is dead. For a discussion about the ambiguity of r^2 for constrained regression, see http://www.graphpad.com/guides/prism...regression.htm