generating an equation for a surface

[#1]

Hi all

I have data where there are two input parameters that yield a factor.  I can make a surface plot of this data.  Is there a way to now see an equation that models this same surface?

Maureen

[#2]

Hi Maureen,

This might be a dumb question, but does the column you are plotting in Surface Plot have a formula in it?  If so, the formula is the mathematical equation for the surface.

[#3]

no formulas are in the three columns.  They are two test results and a factor related to them.  Because the data is individual values, I'd like to have a formula that shows the relationship between them so that it is smooth and continuous.

[#4]

Sounds like you need to build a regression model. The form of the model will depend on what that surface actually looks like. This is the wrong forum to teach you everything you might need to know about model building, but try something like this:

Choose Analyze > Fit Model.

Put your response as the Y.

Highlight your two inputs and click on Macros > Response Surface.

Run this model and save the Prediction Formula. The thought is that the prediction formula will be the equation for your surface. However, without seeing your surface, there is no guarantee that this will match. You may need to specify a different model so that you get the proper form for your surface.

[#5]

Ah. The macro Response Surface is the key.
It creates a quadratic function when it looks like I need a cubic function. Is it possible to specify this?

[#6]

Yes, you can specify any polynomial model you want. To add cubic terms, change the Degree box to be 3, highlight the input variables and click Macros > Polynomial to Degree.

[#7]

An option would be use a neural network model, which would fit a smoother surface to your data.

If the data table is not very large and the random error around the surface is small you could also use the Gaussian Process modeling platform, which applies a different kind of smoothing model.