JMP offers specification proposal?

[#1]

Hi.

Supposing that I have bunch of data but no Specification (LSL and USL).

Data are diagnosed as to follow SHASH or other non-normal distribution.

Is there appropriate analougues for the plus/minus 3 SD for normal distribution? (Since 6SD of normal distribution is usually used as Specification Tolerance)

Alternatively, is there function that JMP calculated the possible Specification which ensures specific % will be preserved within the Tolerance? (for the SHASH or Weibull or other distributed data)

Thank you!

[#2 · Accepted solution]

I won't go to the discussion what should determine your specification limits (there are already good topics regarding this in community) or how they should be calculated, but below are some examples what JMP can do

• Fit distribution's Calculate Quantile Spec Limits Options 
• Design Space Profiler Profilers > Profiler > Design Space Profiler 
• Interactive Capability Plot

[#3]

Look into Tolerance Intervals.  There is a non-parametric option you can use.  Available in Distribution.

[#4]

Also, you can fit the SHASH distribution to the data, then save the transformed values to the data table, that transform the data to look like a normal distribution.  You could then calculate tolerance limits on that transformed data and do the backwards transformation to the original scale of the data.  

[#5]

There is a SHASHInv( ) function you can use in a table column or a JMP script to do the back transformation.