Fit Model - Limiting Model Response to Only Positive Values
May 21, 2020 1:02 PM(85 views)
I have recently created a standard least squares model using the Fit Model tool. I want to use the Simulator tool to simulate about 5000 sample runs of my model. I'm assuming it takes the generated regression equation from my model and just inputs 5000 combinations of my model variables. My original data is unique in the sense that all of my input variables and the response will never be negative (essentially, I will never have input or response values existing outside the first quadrant). I can set limits for the Simulator but I get some of my response predictions are negative.
What do I need to do when setting up the Fit Model to ensure that the regression equation results in only positve value predictions?
I'm wondering if you are seeing negative values because a normal distribution doesn't apply to your data. The Generalized Regression platform in JMP Pro can accommodate for this. If you have standard JMP, you may try using a log transform in the Fit Model dialog (highlight the response and then press the red hotspot beside "Transform").