Nonlinear Regression Analysis

  • Thread starter Thread starter Cougar
  • Start date Start date
C

Cougar

Good afternoon,

I have a fairly large data set (roughly 7,000 samples) on which I
would like to run some nonlinear regression analysis.

I'm fully capable of using the Data Analysis Toolpak and running the
linear analysis there, but unless I'm totally missing it, there
doesn't seem to be a method for running nonlinear analysis.

To be clear, what I want to do is take this data, which is in three
columns, and essentially put together an equation to represent Ax + By
= C.

However, this data does not seem to be all that well represented by a
linear equation, and I believe it would be more precise to use an
exponential or logarithmic model along the lines of A^x + B^y = C.

Can anyone help?

Thank you!
 
Good afternoon,

I have a fairly large data set (roughly 7,000 samples) on which I
would like to run some nonlinear regression analysis.

I'm fully capable of using the Data Analysis Toolpak and running the
linear analysis there, but unless I'm totally missing it, there
doesn't seem to be a method for running nonlinear analysis.

To be clear, what I want to do is take this data, which is in three
columns, and essentially put together an equation to represent Ax + By
= C.

However, this data does not seem to be all that well represented by a
linear equation, and I believe it would be more precise to use an
exponential or logarithmic model along the lines of A^x + B^y = C.

Can anyone help?

Thank you!
Click to expand...

Hello:

It may be too late but if you are still interested there are two ways to
do this.

One way is create an xy scatter graph and then right click the series.
When you do this the pop up menu will have Trendline as an option.
Choose this and you will see the built in options, for transformations.
Choose the option you want, some have an additional choice to make. To
see the resulting equation be sure to check the options to display the
equation on the chart.

The second way is to transform the data yourself before you run the
regression (either with the functions or with the Data Analysis wizard).
To this decide on the transformation, transform the data and then run
the regression on the transformed data.

Suppose you want to run Y^2 = a + b X^2 as the model. Create two new
columns of data, by squaring the Y's and the X's. Then run the
regression on the new columns of data. Excel won't know the difference
and will refer to the resulting equation as Y = a + bX, but of course
you know that the Y and X are really squared values. You can do the same
thing with log's or any other transformation you choose.

Pieter Vandenberg
 
Back
Top