Malcolm pointed out that JMP offers a simple, robust and fast way to calculate Cp/Cpk estimates to baseline the performance of continuous variates that do not follow a normal distribution. One simply fits all 13 continuous distributions available in JMP and then lets JMP use AICc criteria to determine the best-fitting distribution. When the specification limits are defined as Column Properties, JMP automatically returns Cp/Cpk estimates and the Long Term Sigma for the best-fitting distribution.
I ran the example with MFI spec limits of 192 and 198 and a target of 195. In this case, the best-fitting distribution is Johnson SI, and the Cp/Cpk indices are calculated from that distribution.
This approach, just a few steps in JMP, can be particularly valuable to those on a Six Sigma team. When they fit the best distribution, they will more accurately predict the future defect opportunity for the process during the Measure and Control steps of DMAIC, or during similar process improvement methodologies they might employ.