[Stuart Saich]

I have a spreadsheet from excel 2003 in which I curve fitted many fifth order
polynomials. I then use the constants from the polynomial function as a
selection for curve generation.

I now want to add some curves and generate constants from trend lines but am
now using office 2007.

The old curve fitted data works well in 2007.

The new curve fitting trend line system cannot generate a fifth or sixth
order polynomial and the polynomial constants generated for a fourth order
polynomial do not generate the same source curve when inserted into the
polynomial equation.



Seems like the polyniimial solving routine has a problem?

Any ideas

Stu

[Jerry W. Lewis]

The formerly excellent chart polynomial trendline engine was "improved" in
2007 to the point that it is no longer trustworthy. LINEST on the other hand
does seem to work better, and you can fit a polynomial with a formula like
=LINEST(ydata,xdata^{1,2,3,4,5})
Note however that fitting a high order polynomial is a very challenging
numerical problem unless your x data spans a very broad range.

Jerry

"Stuart Saich" wrote:

I have a spreadsheet from excel 2003 in which I curve fitted many fifth order
polynomials. I then use the constants from the polynomial function as a
selection for curve generation.

I now want to add some curves and generate constants from trend lines but am
now using office 2007.

The old curve fitted data works well in 2007.

The new curve fitting trend line system cannot generate a fifth or sixth
order polynomial and the polynomial constants generated for a fourth order
polynomial do not generate the same source curve when inserted into the
polynomial equation.



Seems like the polyniimial solving routine has a problem?

Any ideas

Stu