I use LINEST to derive a multiple regression formula in Excel 2003.

I have to share this with a client who uses Excel 97.

While the results for the gradients (m) are the same, the coefficient of
determination (r2) in Excel 97 is lower than that given in Excel 2003. The
results are 0.87 and 0.97 respectively.

Does anyone know why the two versions give different answers? Which if any
is the more reliable calculation?

Regards

In Excel 2003, LINEST (and some other stat functions) was improved. For
details, including why R2 differs in XL97 and XL2003, see
http://support.microsoft.com/kb/828533
You may wish to see how the chart trendline formulas compare.
best wishes
--
Bernard V Liengme
www.stfx.ca/people/bliengme
remove caps from email

"Philip J Smith" wrote in message
...
I use LINEST to derive a multiple regression formula in Excel 2003.

I have to share this with a client who uses Excel 97.

While the results for the gradients (m) are the same, the coefficient of
determination (r2) in Excel 97 is lower than that given in Excel 2003.
The
results are 0.87 and 0.97 respectively.

Does anyone know why the two versions give different answers? Which if
any
is the more reliable calculation?

Regards

That R2 would noticeably change without the coefficients noticeably changing
is surprising to me, unless you are forcing the intercept to zero, as in
LINEST(ydata,xdata,FALSE,TRUE)

If the intercept was forced to zero, then the R2 from pre-2003 versions of
LINEST is known to be wrong, as discussed in part of the Knowledge Base
article that Bernard referenced.

If the intercept was not forced to zero, then I would be very interested in
seeing the data.

Jerry

"Philip J Smith" wrote:

I use LINEST to derive a multiple regression formula in Excel 2003.

I have to share this with a client who uses Excel 97.

While the results for the gradients (m) are the same, the coefficient of
determination (r2) in Excel 97 is lower than that given in Excel 2003. The
results are 0.87 and 0.97 respectively.

Does anyone know why the two versions give different answers? Which if any
is the more reliable calculation?

Regards