NB. This contribution is also posted on the JASP website.

Author: Martin Hayden, Master Black Belt at LearnSixSigma.co.uk discovers JASP.

If you are practising Six Sigma today, the good news is that you don’t need to buy an expensive software package like Minitab. There is free software available that meets my needs, and I suspect it will meet yours too.

As a user of Minitab for over a decade, I believe it’s a great tool. It did everything I needed and could probably do a lot more. However I was increasingly dissatisfied with the rising cost of maintaining the license, so I began looking for a free software alternative.

JASP is the alternative which I have been testing. It can do everything I used to do in Minitab as a Six Sigma practitioner. It also has excellent data presentation and display tools. It is perfect for my teaching role allowing students to have free access to the same software and to participate live in classroom and online teaching examples.

Visualising Distributions

Let’s start by looking at the graphical display of distributions, one of the first things I usually want to do with any new data.

 

For multiple samples, there are box plots and, a new one for me, raincloud plots.

 

Also, scatter plots with a few of the selectable display options.

Process Quality Control

In the Quality Control module, are all the usual control charts used in Six Sigma analysis: X bar R, X bar S, X mR and the attribute charts.

 

Process Capability analysis is a common requirement in Six Sigma. There is a sometimes bewildering variety of metrics used, but the common standards of Cp, Cpk, Pp and Ppk are all available in JASP. Motorola defined several alternative standards, Sigma level and Z bench, and their claimed equivalence with DPMO metrics. JASP stays clear of these remaining with the classic Process Capability metrics. Minitab includes both standards, which can be confusing.

Complex Data Analysis

In numerical analysis, pleasantly JASP has all the standard hypothesis tests that I frequently use. A t-test, paired t-test, chi-square test and normality test. But let’s look at some of the more complex analyses you might need in a Six Sigma project.

Measurement Systems Analysis (MSA) is something I have needed on several occasions. It is available in the Quality Control module of JASP. This data was taken from a classroom exercise measuring the diameter of pennies using a digital calliper. The specification limits were estimated and the purpose was to demonstrate an inadequate measurement system. It was pleasant to see that the results from JASP were identical to those from running the same MSA in Minitab.

 

The data was collected from 3 operators measuring a randomised sample of 10 pennies with 2 repeats of each measurement. The analysis uses an ANOVA to calculate the components of measurement error. The repeatability and reproducibility components are compared to 6 times the standard deviation and the tolerance width. This is shown in a simple traffic light indicator and the familiar-looking data table.

Modelling

Design of Experiments (DOE) is not a frequent requirement for me, however JASP supports DOE in the Quality Control module. This example shows the results from a Statapult (a wooden catapult and training aid for teaching Design of Experiments) screening experiment developed live in a classroom.

 

It followed a standard Placket and Burman model but used only one repetition as an exercise. The data was collected in a spreadsheet using the Placket and Burman matrix model from a textbook. JASP doesn’t have the Assistant Wizard like in Minitab but it isn’t difficult to set up the appropriate model and upload data into the JASP Data File shown. A screening analysis using the main non-interaction terms and gave the following p values and coefficients. Those p values above 0.1 suggest the terms are non-significant and may be deleted from the model.

The remaining four terms are statistically significant and would be considered for inclusion in the next stage. The table shows the standard numerical presentation of results found in all DOE modelling packages. Here the significant terms are the intercept, D, E, F, and A as indicated by the p values below 0.1 which is the standard rule used in DOE in place of the more usual 0.05.

 

A simpler presentation of the model results is shown in this Pareto Chart. It arranges the factors in order of their coded coefficients, a measure of their effect on the response. You might notice it only shows three factors as significant. That’s because JASP defaults to a 0.05 level of significance. If you want to apply a threshold of 0.1 you need to check the data table. Minitab is a little more flexible here allowing you to select the significance level to apply, but JASP provides precisely the same analysis.

Other Observations & Conclusion

JASP has limited data preparation capabilities but can take data from CSV files. Since I do most of my data preparation in Excel with all the slicing and dicing tools it has, exporting the data to JASP is no hardship. JASP will retain the link to the file and refresh the data should I update the original.

I suspect that I have barely scratched the surface of what JASP is capable of. I haven’t yet delved into most of the modules included with the software. One of these days I will investigate Bayesian statistics and some of the other advanced modelling tools.

This is just an overview, but I think JASP’s capability is excellent for Six Sigma analysis. The results are well presented, and in terms of value for money, I think it beats many other systems.