Revealing Latent Quality Information Hidden Within Inspection Reports of Curtailed Tests

• Activity Number: 474
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Quality and Productivity
• Author(s): Tamar Gadrich*+ and Emil Bashkansky
• Companies: Ort Braude College and Ort Braude College
• Keywords: Latent quality information ; Curtailed tests ; Information mining

Abstract

Information mining statistical tools can be used to uncover the latent quality information hidden within empirical data. The desire to reduce inspection costs often results in curtailed tests, i.e. even though every item is meant to undergo a number of (independent/dependent) tests, failure of the first test terminates further tests/inspection. Thus, the empirical data does not contain information about the item's ability to pass the cancelled tests. This paper proposes altering testing procedures (e.g., changing the order of the tests, detectability levels in lot-by-lot or item-by-item inspection), in order to estimate the theoretical joint probabilities (latent quality information) concerning an item's ability to pass a part of or the entire test. The effectiveness of the proposed procedures is then evaluated using simulated data.