I have fit a series of data using a 5th order polynomial from the trendline
dialog. Using the coefficients displayed in the equation, the curve is not
faithfully reconstituted. Is it possible to export the trendline
coefficients to arbitrary significant figures?

Select the trendline equation on the greaph and format the number of
significant figures

"MrUniverseman" wrote:

I have fit a series of data using a 5th order polynomial from the trendline
dialog. Using the coefficients displayed in the equation, the curve is not
faithfully reconstituted. Is it possible to export the trendline
coefficients to arbitrary significant figures?

If you are absolutely, positively sure a 5th order polynomial is the
correct choice for your data fine, but you may want to check
Trendline coefficients
http://www.tushar-mehta.com/excel/ti...efficients.htm

--
Regards,

Tushar Mehta
www.tushar-mehta.com
Multi-disciplinary business expertise
+ Technology skills
= Optimal solution to your business problem
Recipient Microsoft MVP award 2000-2005

In article ,
says...
I have fit a series of data using a 5th order polynomial from the trendline
dialog. Using the coefficients displayed in the equation, the curve is not
faithfully reconstituted. Is it possible to export the trendline
coefficients to arbitrary significant figures?

BJ has noted that you can select the graph trendline equation and
reformat it. For your purposes I recommend scientific notation with 14
decimal places.

Aside from the question of potentially overfitting the data, which
Tushar addresses on his web page, there is also the question of
numerically fitting the model (Tushar has a heading for this, but has
apparently not writen it yet). Unless your x data spans a very wide
range, it is unlikely that you have adequate information to fit a 5th
order polynomial.

If you have Excel 2002 or earlier, and there are appreciable numeric
differences between the chart trendline estimates and the LINEST
estimates, then the fitting problem is ill-conditioned and the LINEST
results are unreliable (the chart trendline estimates are more robust,
but depending on how ill-conditioned may also be numerically unreliable).

For additional discussion of this topic, see
http://groups.google.com/groups?selm...0no_e-mail.com
which references
http://groups.google.com/groups?selm...enetserver.com

Jerry

MrUniverseman wrote:

I have fit a series of data using a 5th order polynomial from the trendline
dialog. Using the coefficients displayed in the equation, the curve is not
faithfully reconstituted. Is it possible to export the trendline
coefficients to arbitrary significant figures?