Reproducibility statistical tools

This paper explains how to use statistical tools to evaluate the reproducibility of color testing. Several pigment chemistries combined with several test methods were evaluated in hopes of establishing a test that both the statistician and color technician would endorse. 

Pigments and pigment dispersions have been and continue to be evaluated for color using a wide variety of methods. Over the years Oil Ink Tests, Latex Paint Tests, Liquid Ink Tests and PVC, Rubber, and Polyethylene Two-Roll Mill Tests have been used as Quality Control methods. We did not have extensive reproducibility data for these methods, but we felt that all of them could be improved using statistical tools. 

Random daytime (RDT) sampling offers an unbiased approach for zonal compliance assessment but suffers from poor reproducibility, particularly with small sample numbers. Bulking RDT data over several years enables treatment changes to be adequately demonstrated, albeit allowance must be given to the response time of lead pipes and statistical tools can be used to determine levels of confidence in RDT sampling data. 

One approach that QA would use to assure itself that a given process (step) is under control is the effort associated with the concept of process capability. Ekvall and Juran [15] defined the concept as the measured inherent reproducibility of the product turned out by the process. The statistical definition of process capability is that all the measured values fall within a 6-sigma range (i.e., range of the minimum to maximum limits). The information is used to show that the process is under control over a period of time as well as determine whether there is any drifting or abnormal behavior from time to time. Process validation is a QA tool in this case because its data will be used as the origin for the data curve developed for the process capability concept. 

The crisis of the GME method is closely related to the crisis in the density matrix approach to wave-function collapse. We shall see that in the Poisson case the processes making the statistical density matrix become diagonal in the basis set of the measured variable and can be safely interpreted as generators of wave function collapse, thereby justifying the widely accepted conviction that quantum mechanics does not need either correction or generalization. In the non-Poisson case, this equivalence is lost, and, while the CTRW perspective yields correct results, no theoretical tool, based on density, exists yet to make the time evolution of a contracted Liouville equation, classical or quantum, reproduce them. 

Gauge repeatability and reproducibility study Brainstorming and the traditional seven tools of quality Problem-solving methods The seven tools of management Failure mode and effect analysis Statistical process control Control charting... 

---