Using linest or logest for non-linear equations
 

I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

Wow, I don't know my trig anymore.
I don't suppose you could fit an explanation of that equiation in 100 words
or less?

" wrote:

I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

So we need to find the a,b,c and d coefficients. A is a constant, b is the
intercept, c is the coefficient to exp(-c*x) and d is the power of x. It is
known as the Weibull function. Can linest only be used for the specific
power, log, linear, and polynomial equations?

"Sean Timmons" wrote:

Wow, I don't know my trig anymore.
I don't suppose you could fit an explanation of that equiation in 100 words
or less?

" wrote:

I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

Maybe look at the TREND function, since you know your Xs and Ys?
--
Brevity is the soul of wit.


" wrote:

I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

Google is our friend here, since you named the function:
http://www.qualitydigest.com/jan99/html/weibull.html

Dave
--
Brevity is the soul of wit.


" wrote:

So we need to find the a,b,c and d coefficients. A is a constant, b is the
intercept, c is the coefficient to exp(-c*x) and d is the power of x. It is
known as the Weibull function. Can linest only be used for the specific
power, log, linear, and polynomial equations?

"Sean Timmons" wrote:

Wow, I don't know my trig anymore.
I don't suppose you could fit an explanation of that equiation in 100 words
or less?

" wrote:

I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

I would think very hard about it, then try to design a Solver approach. The
idea is to minimize error of the predicted relationship: vary the four
parameters a-d while minimizing the difference between x-yActual pairs and
x-yPredicted pairs.

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


"
m wrote in message
...
I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I
have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since
my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

I treid using the solver approach and could not get the correct values.
There is no way to manipulate the equation using the linest function?

"Jon Peltier" wrote:

I would think very hard about it, then try to design a Solver approach. The
idea is to minimize error of the predicted relationship: vary the four
parameters a-d while minimizing the difference between x-yActual pairs and
x-yPredicted pairs.

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


"
m wrote in message
...
I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I
have
the x and y values but want to smooth the data. I've been trying to fit a
non-linear trend to the equation using the power and log trends but since
my
equation differs I cannot get them to work. Do you know any way to find
these coefficients?

Using linest or logest for non-linear equations
 

Linest has no facility for estimating so many convoluted parameters. If you
transform your data, you can use LINEST to compute Weibull parameters:

http://www.qualitydigest.com/jan99/html/weibull.html
http://www.duxbury.com/authors/varde...bull_plot.html

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


"
m wrote in message
...
I treid using the solver approach and could not get the correct values.
There is no way to manipulate the equation using the linest function?

"Jon Peltier" wrote:

I would think very hard about it, then try to design a Solver approach.
The
idea is to minimize error of the predicted relationship: vary the four
parameters a-d while minimizing the difference between x-yActual pairs
and
x-yPredicted pairs.

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


"
m wrote in message
...
I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I
have
the x and y values but want to smooth the data. I've been trying to
fit a
non-linear trend to the equation using the power and log trends but
since
my
equation differs I cannot get them to work. Do you know any way to
find
these coefficients?

Using linest or logest for non-linear equations
 

Thanks for your help. I'm starting to think that this might not be possible
without using solver. However, when I use solver, the values are not
dependable because solver needs initial parameters. If anyone else has any
ideas please let me know.

Thanks!

"Jon Peltier" wrote:

Linest has no facility for estimating so many convoluted parameters. If you
transform your data, you can use LINEST to compute Weibull parameters:

http://www.qualitydigest.com/jan99/html/weibull.html
http://www.duxbury.com/authors/varde...bull_plot.html

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


"
m wrote in message
...
I treid using the solver approach and could not get the correct values.
There is no way to manipulate the equation using the linest function?

"Jon Peltier" wrote:

I would think very hard about it, then try to design a Solver approach.
The
idea is to minimize error of the predicted relationship: vary the four
parameters a-d while minimizing the difference between x-yActual pairs
and
x-yPredicted pairs.

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


"
m wrote in message
...
I need to find four coeffecients for the equation y=a-b*exp(-c*x^d). I
have
the x and y values but want to smooth the data. I've been trying to
fit a
non-linear trend to the equation using the power and log trends but
since
my
equation differs I cannot get them to work. Do you know any way to
find
these coefficients?

Using linest or logest for non-linear equations
 

y=a-b*exp(-c*x^d).
I have the x and y values
However, when I use solver, the values are not
dependable because solver needs initial parameters.

Hi. If interested, I got Solver to work with about 30 test points.
I don't have your data, but just note that x must be =0.
For example, with your x^d, if we had a neg x, and d was 3.4:
=POWER(-1.2, 3.4)
then Excel returns #Num! error, as it can't return the Complex number.
On my test data, a,b,c,& d were given small positive values (~1-3)

In general, Solver can not do Least Square fitting.
The nature of your equation may allow the following technique.
Have your column of x & y values (name y values "y")
Have 4 changing cells names a,b,c_, d. (Note name "c" must be "c_" in
Excel)
Initially put something like 1 in each cell.
Next to y, put your equation based on the changing cells. (name it something
like fx)
=a-b*EXP(-c_*x^d)
Next column is the error =y-fx .
Target cell is the Sum of the errors.

Minimize Target
Changing cells: a,b,c_,d
st:
Target =0
fx<y

Check Option "Assume non-Negative"

Again, this should work, but may not work with your particular values of
a,b,c_ & d.
I would be interested if you emailed me your values.
Thanks.
Dana

--
HTH :)
Dana DeLouis
Windows XP & Office 2003


"
m wrote in message
...
Thanks for your help. I'm starting to think that this might not be
possible
without using solver. However, when I use solver, the values are not
dependable because solver needs initial parameters. If anyone else has
any
ideas please let me know.

Thanks!

"Jon Peltier" wrote:

Linest has no facility for estimating so many convoluted parameters. If
you
transform your data, you can use LINEST to compute Weibull parameters:

http://www.qualitydigest.com/jan99/html/weibull.html

http://www.duxbury.com/authors/varde...bull_plot.html

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


"
m wrote in message
...
I treid using the solver approach and could not get the correct values.
There is no way to manipulate the equation using the linest function?

"Jon Peltier" wrote:

I would think very hard about it, then try to design a Solver
approach.
The
idea is to minimize error of the predicted relationship: vary the four
parameters a-d while minimizing the difference between x-yActual pairs
and
x-yPredicted pairs.

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


"
m wrote in message
...
I need to find four coeffecients for the equation y=a-b*exp(-c*x^d).
I
have
the x and y values but want to smooth the data. I've been trying to
fit a
non-linear trend to the equation using the power and log trends but
since
my
equation differs I cannot get them to work. Do you know any way to
find
these coefficients?