T

#### tommoulds

I am having real trouble with some simple linear regressions in

Excel. I would be excellent if someone could help.

I have various sets of data (one independent variable) that I am

trying to calculate R-squared for, with the Y intercept forced to 0

and without. When i dont have the intercept forced through zero all

my results seem to be acceptable. When I force crossing zero all my R-

squared values are very near 1. I would expect them to all get

slightly worse (most of the data is positively linearly correlated.)

For example for one set of data I get 0.522 for my R-Squared from

Linest with const as true (not forcing zero.) When I change the const

to zero (force through zero) I get 0.984. This is clearly wrong. I

think it should be about 0.49.

Has anyone else seen this? I am aware that the regression stats on

graphs in excel are incorrect but I though LINEST was ok. I am using

Excel 2003.

Thanks

Tom