Hi,

I have data which generates an s-shaped curve. The best trendlines that can
be obtained in Excel are of the order 6 and that isn't close to the accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of order
higher than 6?

Thanks in advance,
Vikram

Do you have a physical basis for a poly fit of such high order? Or do you
want the line just to look nice?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
http://PeltierTech.com
_______


"Vikram" wrote in message
...
Hi,

I have data which generates an s-shaped curve. The best trendlines that
can
be obtained in Excel are of the order 6 and that isn't close to the
accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of order
higher than 6?

Thanks in advance,
Vikram

Thank you for the reply.

There is no physical basis as such but I am looking at data that has to
undergo several sets of such curve fitting combined with inter- and
extrapolation. So in order for the final output to be free of any gross
errors I need to make sure that the error induced by considering these
approximations is avoided. I understand that there is a possibility of
opposing actions of errors (thereby self-compensating), but I would like to
avoid such errors in the first place.

Because I couldn't find a higher order polynomial function , I had divided
the domain into a few groups and am using different functions for different
regions. This one works better for me.

Thank you,
Vikram

"Jon Peltier" wrote:

Do you have a physical basis for a poly fit of such high order? Or do you
want the line just to look nice?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
http://PeltierTech.com
_______


"Vikram" wrote in message
...
Hi,

I have data which generates an s-shaped curve. The best trendlines that
can
be obtained in Excel are of the order 6 and that isn't close to the
accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of order
higher than 6?

Thanks in advance,
Vikram

Which leaves one wondering what the purpose of the fit is for?
--
Bernard V Liengme
www.stfx.ca/people/bliengme
remove caps from email

"Vikram" wrote in message
...
Thank you for the reply.

There is no physical basis as such but I am looking at data that has to
undergo several sets of such curve fitting combined with inter- and
extrapolation. So in order for the final output to be free of any gross
errors I need to make sure that the error induced by considering these
approximations is avoided. I understand that there is a possibility of
opposing actions of errors (thereby self-compensating), but I would like
to
avoid such errors in the first place.

Because I couldn't find a higher order polynomial function , I had divided
the domain into a few groups and am using different functions for
different
regions. This one works better for me.

Thank you,
Vikram

"Jon Peltier" wrote:

Do you have a physical basis for a poly fit of such high order? Or do you
want the line just to look nice?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
http://PeltierTech.com
_______


"Vikram" wrote in message
...
Hi,

I have data which generates an s-shaped curve. The best trendlines that
can
be obtained in Excel are of the order 6 and that isn't close to the
accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of
order
higher than 6?

Thanks in advance,
Vikram

Bernard -

I think they want it to "look good" as a reassurance that any extrapolations
also "look good".

Vikram -

Poly fits aren't bad if you need to interpolate values, even if there is no
basis for selecting such a fit. However, you must check that the fit does
what you want, without dips and bumps everywhere. If you need to
extrapolate, you'd be unwise to depend on a polynomial relationship.

Breaking up a set of data is probably not a bad idea, but you should still
take care with the data.

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
http://PeltierTech.com
_______


"Bernard Liengme" wrote in message
...
Which leaves one wondering what the purpose of the fit is for?
--
Bernard V Liengme
www.stfx.ca/people/bliengme
remove caps from email

"Vikram" wrote in message
...
Thank you for the reply.

There is no physical basis as such but I am looking at data that has to
undergo several sets of such curve fitting combined with inter- and
extrapolation. So in order for the final output to be free of any gross
errors I need to make sure that the error induced by considering these
approximations is avoided. I understand that there is a possibility of
opposing actions of errors (thereby self-compensating), but I would like
to
avoid such errors in the first place.

Because I couldn't find a higher order polynomial function , I had
divided
the domain into a few groups and am using different functions for
different
regions. This one works better for me.

Thank you,
Vikram

"Jon Peltier" wrote:

Do you have a physical basis for a poly fit of such high order? Or do
you
want the line just to look nice?

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
http://PeltierTech.com
_______


"Vikram" wrote in message
...
Hi,

I have data which generates an s-shaped curve. The best trendlines
that
can
be obtained in Excel are of the order 6 and that isn't close to the
accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of
order
higher than 6?

Thanks in advance,
Vikram

Polynomials have a fundamentally different shape than s-shaped curves, since
polynomials are unbounded and s-shaped curves are not. If you are fitting a
broad enough range that you need a 6+ degree polynomial, you might do better
fitting a function that is actually s-shaped. Without knowing what kind of
data you have, the best I can suggest is that a 4- or 5-parameter logistic
(Google it) is used frequently in biochemistry. Any cumulative probability
distribution function is also s-shaped.

"Vikram" wrote:

Hi,

I have data which generates an s-shaped curve. The best trendlines that can
be obtained in Excel are of the order 6 and that isn't close to the accuracy
I am seeking.

Does anyone know if there is a way in Excel to get a trend line of order
higher than 6?

Thanks in advance,
Vikram

Thanks to everyone for answering my question. I am only looking to
interpolate between points. So I guess dividing the domain and then trying to
fit different functions to these disparate regions would actually give better
predictions within that division.

Thanks again,
Vikram