In the documentation Simulation Settings for Responses (jmp.com), we can see that the option "Random by Model" is active for Least Mean Square and Genreg models and that it adds an error based on the model distribution specified in the fit model launch window.
it's very clear to me in the genreg situation, and I've tested it with a lognormal distribution and it works very well.
What about Standard Least Square? do we have situation where the distribution is différent than Gaussian Disitribution?
is it linked to box-cox transformation option? https://www.jmp.com/support/help/en/16.2/?os=win&source=application#page/jmp/box-cox-y-transformation.shtml
What about the log variance? https://www.jmp.com/support/help/en/16.2/?os=win&source=application#page/jmp/overview-of-the-loglinear-variance-model.shtml
What about weight? Is it not used as we can't predict it for future values or is it used if the weight column has a formula which can help to define the weight of future prediction?
Thanks a lot,
Florence
