Hi JMP community,
(JMP Pro 16.0.0)
I'm hoping to find out if there is a way to include lower bounds to simulated data in the Profiler platform. I have several predictors going into a model, but few data points to fit. So, in order to generate simulated ("synthetic") data, I use the Prediction Profiler platform along with the multivariate correlation structure of the data to generate thousands of simulated data points that maintains the same correlation structure as the original data. See example screenshot of the simulator functionality of the profiler.
One of the issues I'm facing is that I have a couple predictors whose real values can never be less than zero. However, when setting the values to generate the simulated data, JMP will use the mean and standard deviation along with the correlation structure to create the simulated data. For those predictors whose values can never be less than zero, JMP creates some data points where that is the case.
I know I can manually go in and select the rows where there are negative values for the predictors, and can even automate this with JSL, but I was hoping to see if there's a way to have this be part of the settings of the simulator so that when it generates the data, it takes into account that some (or even all) predictors can't have values below a certain limit (zero in my case).
I have to maintain the correlation structure from the original data, so I can't select "Random" as the distribution type, I have to select Multivariate. Also, if I fit the distributions of the real data, the best fits are SHASH curves, which aren't an option for the "Random" distribution type in the simulator.
I don't think there is a possibility to do what I want, and if it's confirmed, then I might submit it as a wish list for future versions of JMP.
Thanks!,
DS