Even if the "how-to" descriptions are short click instructions, a little "author's philosophy" must be added beforehand, especially for the attributive and ordinal procedures.
In general, there is quite a large and comprehensive documentation for this:
Link to solara.MP Attributive Analyses
Attributive analysis has changed massively over the years. Just 20 years ago, X components were tested X times by X operators. They all had to say the same thing. And the result was a green smiley.
However, this had one major disadvantage (and the author is now using the same words here that he would use in discussions with customers): a well-trained orangutan could have carried out such an "MSA" if it had simply been presented with various "perfect components" and various "extremely bad components". However, the validity of such an analysis is then open to debate.
Methods therefore had to be developed to cover the "borderline case". To get a real statement about the quality of the "inspection".
The effort, especially the effort involved in taking samples, has become massive here. However, one thing must be considered: In the manufacturing process, the aim is to “inspect” instead of measuring. Faster and cheaper. And this requires more effort when analyzing.
A distinction must be made between 2 basic methods:
- MSA attributive: These are "discretized characteristics". Characteristics that could actually have "continuous" measured values, but are to be analyzed attributively. (only “inspected”)
- MSA ordinal/nominal: real "discrete characteristics".
In addition, a distinction is often made between "Is there a "reference existing" or not?" The author considers this question to be invalid. There is always a reference, even if the decision is made by the master, the Quality management or a higher organization. A reference is always available.
The "how-to" descriptions are therefore reduced to one example of each of these two types, always with reference.
- MSA attributive: Calculation of the Signal detection
- MSA ordinal/nominal: Effectiveness analysis
Other options are available, such as Kappa or Bowker test, but their application and interpretation must be learnt in training courses.