Change a distribution of bin outcomes into poisson representing profitability

Question

Hi folks, I apologise for my level of unfamiliarity. I'm diving through the manuals trying to discover the tools I need, and I'm positive what I'm trying to do must be easy. But I'm just new and struggling a little bit.

I have here some data representing different bin classes of a manufacturing process, a metric scoring the value of each outcome. In this particular process each bin class has equal probability of occurring.

What I want to do is turn this into a 2nd order distribution representing the likely value of a larger run. Say, I want to take 10 results from the list using simple random sampling with replacement. Now, I will sum all the chosen results. Call this distribution G{k}, the profitability for a run of k units.

How do I turn my first order distribution into this second order distribution using jmp? I'm a bit lost in all the documentation. Like I said I'm happy to read, so if you point me to a tool within JMP or some function in the macro language I'll be able to hunt it down in the manuals.

Thanks for any guidance!

avancouw · Answer

er what i meant to say: I want to develop the cumulative probability function (graphically is good enough) for this second order distribution G.

avancouw · Answer

Thank you so much for the answer. I recall this term bootstrap simulation from statistics class, and I think it'll turn out to be what I need. My memory is a bit fuzzy. I'm going to have to use this toolset more so that I become more familiar.

Anyway I'll go get to reading and let you know how it turned out!

Thanks,

avancouw · Answer

@pauldeen wrote:What is the goal behind the request? Are you trying to calculate some percent above a certain value (Percent out of spec)?

I want to see visually a cumulative distribution graph for a run of k product, to determine how large runs must become before we can count on profitability more than ~90% of the time. I will vary k until 90% or more of outcomes yield more value than their input cost.

I expect to need k around 10 but it would be better to develop some evidence rather than operating on a guess.