Hi all, could someone please kindly advice if Fit Model in JMP can be used for data that are not normally distributed (non-parametric)? I look forward to hearing anyone's thoughts or advice on this. Thanks a lot!
Using the distribution platform and fitting all distribution one can get the best fit for the data. Using the Fit Model platform you can leverage the Box-Cox transformation and fit the transformed data or you can look under the red triangle in the lower region of the Fit Model dialog box and look at the transformations available to you. Hope this helps
If you are looking for the commonly accepted nonparametric regressions like kernel regression and quantile regression, then Fit Model does not support it. However, JMP has a SAS Add-In that may help you. In JMP, select File>SAS>SAS Add-Ins. Of interest to you might be the Loess Regression, Quantile Regression, and Thin Plate Spline procedures. These are good when you don't know a suitable parametric form of the response. Click on the Add-In Help link for details. To use these, you need SAS.
Fit Y by X has a Fit Spline feature, but is only used for a single X. For multiple X's, you can use the Knotted Spline Effect attribute in Fit Model. See the documentation for details on both of these.
The Neural and Gaussian Process platforms are also useful for fitting very flexible regressions models.