I'm working a JMP partner who is preparing to show the Design Space Profiler (DSP) at an upcoming conference. Need assistance with clarifying the two questions below:
1) The random uniform distribution used by DSP has no random seed setting so every launch of DSP will produce different inspec portion & volume portion results, is this true or is there a way to obtain repeatable results?
2) Why is DSP not available from the profiler for tree-based models? I believe it's because a tree based model could produce a very narrow range of factor settings for predictions which would make getting a good estimate difficult, is this the reason?
Thanks for your assistance!
