Edit: After messing around with all the settings, I think I've finally got the answers I was looking for. I'm keeping the post but added the comments below for those who might be wondering the same things.
Hi all,
I've been playing around with the Simulator features in JMP and have a few things I'd like to clarify. Attached is an example from the Fitness sample data set.
Under the Profiler plots, I specify the distributions for the 3 model parameters. The histogram on the right shows the distribution of the response after 100 runs with the input parameters varied according to the specified distributions.
Under the Simulate to Table > Sequencing, I select "Sequence Location" as the "Sequence" and select the range of each of the input parameter mean. "Number of Steps" specifies the number of means to be tested. Now, when I click "Make Table", it will simulate a total of 5x5x5x100 runs. For each of the 5x5x5 combinations of the input parameter means, 100 runs will be simulated. My first question is, are the SDs for each set of 100 runs be the same ones that I specify under the histograms earlier? (Yes) Looking at the generated table, I believe this to be the case, just wanted to clarify.
Now, when I select "Simulation Experiment", I'm given the options to select which input parameters to be used for the simulation. If I only select "Runtime" to be simulated, does that mean "RunPulse" and "MaxPulse" will be varied according to the distributions specified earlier in the histograms while "Runtime" will be fixed at a few levels depending on the range shown in the Profiler and the number of experimental runs? Or would "Runtime" at each experimental run be varied around the set level with the previously specified SD? (At each experimental run, "Runtime" will be varied around each set level/mean based on SD specified in Prediction Profiler. Also, the range of "Runtime" means to be tested is set in the Prediction Profiler plot, not under the Simulate to Table boxes)
For the former, I imagine The simulation would go like below, if I set number of experimental runs to be 3 and full factor range:
Exp Run1:
N Run1: Runtime = Random Normal (8, 1.172); RunPulse & MaxPulse vary according to histograms
N Run2: Runtime = Random Normal (8, 1.172); RunPulse & MaxPulse vary according to histograms...
N Run100: Runtime = Random Normal (8, 1.172); RunPulse & MaxPulse vary according to histograms
Calculate defect rate from 100 runs
Exp Run2:
N Run1: Runtime = Random Normal (11, 1.172); RunPulse & MaxPulse vary according to histograms...
N Run100: Runtime = Random Normal (11, 1.172); RunPulse & MaxPulse vary according to histograms
Calculate defect rate from 100 runs
Exp Run3:
N Run1: Runtime = Random Normal (14, 1.172); RunPulse & MaxPulse vary according to histograms...
N Run100: Runtime = Random Normal (14, 1.172); RunPulse & MaxPulse vary according to histograms
Calculate defect rate from 100 runs
I feel like this wouldn't make a lot of sense because if all 3 parameters are selected for the Simulation Experiment, wouldn't the defect rate be either 0 or 1 since all N Runs for each experimental run would yield the same result (given no noise is added to the response)?
Thank you!
