Hi All,
I'm hoping someone might be able to help me to understand what is causing some strange behavior when I use GenReg and the Forward Selection Estimation Method.
So, I have some data that I'm running through GenReg with Forward Selection, where I have several model effects and a validation column as well. After performing the fit, I want to run some simulations on the Estimate column of the report, so I first save a Simulate Response column to the data table. When I do this, my first 614 rows all have very strange values -- values that are essentially 0 with a standard deviation of about 4.5. But then, from row 615 on, the data in the formula column suddenly jumps to values close to what it should be modeling, around 130 with a standard deviation of about 10. I do not understand how or why this is happening.
If I save the Prediction Formula to the data table, it also shows a strange behavior. For those first 614 rows there are no values in the Prediction Formula column, and then there are values from row 615 on, see data table screenshot below and of the distributions.
If I test out the other Estimation Methods, I get a similar behavior with Standard Lease Squares, and Ridge. The other methods don't seem to exhibit this strange behavior.
Does anyone know the cause of this and how to fix it? Or, if you've experienced something similar, how were you able to overcome this problem and have JMP correctly save the Simulation Formula?
Thanks!,
DS
Update/edit: I was able to reproduce the same issue with Pruned Forward Selection as well, but it doesn't happen every time.