Avoid using excluded rows as the validation portion of the model

Peter_Hersh

I have a data set that I have gone through and excluded outliers.  Then I build a model with that data.  By default JMP uses the excluded rows as validation.  Does anyone know of a way to turn this off?  I don't want the outliers to be the validation portion of the model.  I know you can create a separate validation column and use that instead in Pro, but in standard JMP that is not an option.  Anyone have input on this it will be greatly appreciated.

Thanks,

Peter Hersh

DonMcCormack

Pete,

You have two options (three if you include buying JMP Pro). First, you can use a weight column and give all the excluded observations 0 weight. This works if the problem is with the response, but will remove the entire row in cases where only one or a few of the predictors are the problem. The other option is to use Missing Value coding. More tedious, particularly if you have lots of columns, but selectively removes the observations on a cell by cell basis.

Don