nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

is there something like the curve fit platform where you can try different transformations of data to find the best linear fit?

MaureenS

On the curve fit platform you can try different types of curves and see which one is the best fit. I would like to be able to compare different data transformations to find the best linear fit. Is there a platform that can do this?

I am using JMP 17.0 on a MacBook Pro.

Accepted answers

All comments

txnelson

Look at

Analyze=>Specialized Modeling=>Fit Curve

Look for the documentation at here

Help=>JMP Documentation Library=>Predictive and Specialized Modeling

MaureenS

but I only want a linear fit. I want to try different data transformations to see which gives the best linear fit.

txnelson

An easy way to do that would be to create a separate new column for each of the transformations you want to use, and then use Fit Y by X platform and generate the Fit Line. You can then order the different regressions by Goodness of Fit

dlehman1

I think you need to distinguish between linear functions and linear models. Polynomials are nonlinear functions but are linear models - Y=f(X,X2, X3, etc.) can be a linear function. Fit curve will fit such polynomials. It will also fit things that are not linear models. So, you can confine yourself to the polynomials for a start. Now, if you want logarithmic transformations, then I think you will need to follow txnelson's advice and create columns for these. Of course, a log transformation is a nonlinear function, but you can have a linear model in the logs of X and/or Y. No platform will automatically cover all potential transformations. I don't think you are insisting on only linear functions - least square regression will give you the best fitting linear function and there is no need to try others since the regression line (surface) is already the best fitting one.

hogi

If you want to apply a Box Cox transformation before fitting the data, you can use the Fit Model platform and in the red triangle menu select

Then go to the Box-Cox menu and select

and you will get a fit for the transformed variable:

Mark_Bailey

JMP provides an automated Box-Cox Transformation for the response in Fit Least Squares. This method generalizes the common transforms such as logarithms or roots.