Distribution Nonconformance Statistics vs. Distribution Profiler

[#1]

Hello JMP Community,

I have a set of data that is best fit with a lognormal distribution. I cannot seem to find what is going on behind the scenes for the Nonconformance statistics vs. the 95% confidence intervals of the distribution profiler. When I run the capability analysis, I get a nonconformance table. Observed % is obvious in that my actual data population did not have any values below the LCL. What statistics are used for calculating the Expected Overall %, I assume it is using the lognormal distribution and making a judgement as to how well it fits the data; is it a 3 sigma approach or something? How is it calculated, and what useful information does it provide as opposed to the confidence intervals of the distribution profiler. 

The confidence intervals I think are interpreted in the following manner: since I only have a LCL, I would look at the upper 95% confidence interval to make a statement along the lines of 'with 95% confidence one can expect 0.91% of values to fall below the LCL.

Thank you for any help on this!

Greg

[#2 · Accepted solution]

The difference between Observed and Expected is that Observed looks at your data (in the table) and Expected calculates % of LogNormal Curve below LSL. So one is describing the sample and one is predicting HVM from the fitted model.

Capability analysis answers the question: How good is this product at meeting Spec limits!

The CI on the curve of predicition profiler says something about the certainty of your predicted probaility (of whatever your response is). So at 50000 Impedance/2 you predict a probability of 0.0003 (based on the sample) with a 95% certainty of the HVM value being somewhere between  4.4e-6 and 0.009.

So in short they are two very different pieces of information!