Visual Attribute Inspector Capability and KAPPA

[RLKESTER]

[FONT=&quot]Seeking assistance regarding properly statistical method for determining the quality of a lot of product for a specific visual attribute; the inspector was not documented as being qualified to perform the visual attribute inspection.

As a retrospective effort the desire is to establish that the inspector was qualified by conducing a retrospective Qualification of the inspector.

Several methods have been proposed:
1) KAPPA method demonstrating multiple inspectors ability to agree on inspection results with a confidence of 0.8[/FONT]
[FONT=&quot]
2) Apply ISO 2859-1:1999 Sampling Procedures For Inspection By Attribute and based on the lot size and risk assessment establish an AQL; then with a known population of parts (Pass/Fail) conduct a blind study allowing the inspector to perform an inspection and thereby infer an outcome.[/FONT]
[FONT=&quot]
[/FONT]
[FONT=&quot]Your thoughts specific to the appropriateness of KAPPA vs [/FONT][FONT=&quot][FONT=&quot] ISO 2859-1 [/FONT]are appreciated.[/FONT]

 

[Jen Kirley]

A quick bump: can anyone answer this question?

 

[Bev D]

A kappa study is more appropriate than the second choice. The second choice will be affected by the sampling in addition to the inspector's ability. The Kappa approach is generally accepted as an appropriate method.

I will caution you that a Kappa score of 0.8 is terrible for an inspector. The Kappa score should be greater than .9, preferably greater than .95.

You can see my attachment on how to perform a kappa for inspector qualification in this thread, post #7. There is a section on the Kappa test included in this article.

If you need a spreadsheet to perform the calculations you will find it here.

 

[RSS123]

good info