Hi

I have plotted a trend line based on some actual known data. I have
then manually calculated the y figure by plugging in the trend line
formula.

y = 5E-05x4 - 0.0283x3 + 5.4408x2 - 521.05x + 29719

On that attachment you can see a difference in results even though its
the same trend line.

Any ideas?

Thanks

Wired


+-------------------------------------------------------------------+
|Filename: dummy data.gif |
|Download: http://www.excelforum.com/attachment.php?postid=5103 |
+-------------------------------------------------------------------+

--
Wired PSF
------------------------------------------------------------------------
Wired PSF's Profile: http://www.excelforum.com/member.php...o&userid=36795
View this thread: http://www.excelforum.com/showthread...hreadid=565052

Wired PSF -

(1) Use considerably more significant digits (up to 15) in your
calculations. The first term, 0.00005*x^4, lacks precision.

(2) Avoid overfitting your data. A polynomial of order 4 is seldom
appropriate.

- Mike
http://www.mikemiddleton.com


"Wired PSF" wrote
in message ...

Hi

I have plotted a trend line based on some actual known data. I have
then manually calculated the y figure by plugging in the trend line
formula.

y = 5E-05x4 - 0.0283x3 + 5.4408x2 - 521.05x + 29719

On that attachment you can see a difference in results even though its
the same trend line.

Any ideas?

Thanks

Wired


+-------------------------------------------------------------------+
|Filename: dummy data.gif |
|Download: http://www.excelforum.com/attachment.php?postid=5103 |
+-------------------------------------------------------------------+

--
Wired PSF
------------------------------------------------------------------------
Wired PSF's Profile:
http://www.excelforum.com/member.php...o&userid=36795
View this thread: http://www.excelforum.com/showthread...hreadid=565052

Also try to check the t-values of each coefficients.

Try to use LINEST function.

Note: LINEST is a matrix function. You need to enter the formula by
ctrl-shift-enter rather than just enter. See the help function from
LINEST. It will tell you more about multivariable regressions.

Also note that office2k and office may give slightly different results if
your variables are inter-related.

Man







"Mike Middleton" wrote in message
...
Wired PSF -

(1) Use considerably more significant digits (up to 15) in your
calculations. The first term, 0.00005*x^4, lacks precision.

(2) Avoid overfitting your data. A polynomial of order 4 is seldom
appropriate.

- Mike
http://www.mikemiddleton.com


"Wired PSF" wrote
in message ...

Hi

I have plotted a trend line based on some actual known data. I have
then manually calculated the y figure by plugging in the trend line
formula.

y = 5E-05x4 - 0.0283x3 + 5.4408x2 - 521.05x + 29719

On that attachment you can see a difference in results even though its
the same trend line.

Any ideas?

Thanks

Wired


+-------------------------------------------------------------------+
|Filename: dummy data.gif |
|Download: http://www.excelforum.com/attachment.php?postid=5103 |
+-------------------------------------------------------------------+

--
Wired PSF
------------------------------------------------------------------------
Wired PSF's Profile:
http://www.excelforum.com/member.php...o&userid=36795
View this thread:
http://www.excelforum.com/showthread...hreadid=565052