We’ve all heard the proverb, Measure Twice - Cut Once used figuratively to mean Plan Thoroughly Before You Act. Taken literally, of course, it means we should double-check measurements for accuracy before using them to cut (or destroy) something.
But what if you use a different ruler, or have someone else, like a child, take one of the measurements? Perhaps one ruler measures to the ¼ inch, the other to the ⅛ inch. Or suppose you are measuring in different environments (under water or in or desert) that might impact the ruler’s material or markings. How do you decide where to cut?
JMP users know how to plan. They know how to measure. And, they know the importance of understanding the factors impacting their measurement systems (their rulers) – including tool precision, operator characteristics and environmental impact on their measurement tools. So how do they integrate all this knowledge? Measurement Systems Analysis (MSA).
JMP Research Statistician, Caleb King, is one of JMP’s DOE (Design of Experiments) software developers. One of his recent contributions is to the JMP 17 MSA Design tool that helps users easily design and evaluate MSA studies prior to implementation.
Caleb explained that MSA studies are a unique type of analysis where the focus is on identifying variance components rather than treatment effects. For example, a manufacturer of capacitors may want to know how much of the variation in their product’s capacitance is due to the tools used to measure capacitance rather than the natural part-to-part variation.
Caleb summarized what’s new:
Special Purpose DOE. A new Special Purpose DOE (Design of Experiments) option lets users specify factors, factor levels and a randomization scheme in an MSA. Then, JMP provides design diagnostic measures for evaluating the designs. Users can select from three replication schemes (fast repeat, batch repeat and completely randomized) that JMP uses to create an MSA design table and scripts to run analyses after data collection.
Nested and/or Crossed Factors. Nested factors are common in MSA studies. Such factors have levels that are specific to the levels of one or more other factors. For example, suppose a capacitor manufacturer has two plants, each with its own set of operators responsible for measuring capacitance. The operator factor is then nested within the plant factor because the individual operators are specific to each plant. If all the operators were forced to use the same set of tools for measuring capacitance, then the tool factor is then crossed with the operator and plant factors.
In MSA Design, users can access an Interactive Nesting Structure feature to create a custom design structure containing nested and/or crossed factors by picking the factor type for each factor.
Assessing Design Performance. To the best of our knowledge, evaluation of MSA designs has not yet been considered in the literature or other software. So, we now provide MSA Design capabilities that are not currently available elsewhere and we’ve built them to be consistent with, and as full featured as, our other design tools.
In MSA Design, users simulate planning values to assess design performance and Gauge R&R and EMP metrics across a range of conditions. Users can also adjust the number of factor levels/replicates and update the design, if desired, based on the improved diagnostics. Finally, users can export the simulated variances to construct their own metrics.
Caleb talked to me about his background, how he approaches software development and shared a bit about his other interests.
Want to learn about MSA Design using JMP 17? See video and JMP journal/files from Caleb's Oct. 27, 2022 Designer Tutorial: Designing and Evaluating Measurement Systems Studies Prior to Implementation
