Attribute Gage Study (Analytic Method) - Introductory Overview

Attribute gage studies are conducted in order to assess the amount of bias and repeatability in a gage when the response is a binary (such as accept or reject) attribute variable.

In Statistica , two methods for testing bias are available:

  • the AIAG method
  • the Regression method

Each method, involves obtaining the reference value for 8 parts, measuring those parts a number of times (15 or more times for the regression method, exactly 20 times for the AIAG method), and recording the number of acceptances. Once this step is completed, the results can be used to determine the amount of bias and repeatability for the gage.

Part selection

The selection of parts for the study is important. According to the AIAG (2002), you need to have eight parts:

  • one part that is always rejected,
  • one part that is always accepted
  • six parts that have a combination of acceptances and rejections

Reference values for these parts need to be as equidistant as possible with the smallest and largest reference values representing the minimum and maximum of the process range. When using the regression method (see the Computational Details), you can have more than six parts between the two extreme reference values; however a minimum of eight parts is always necessary.

Number of trials

As indicated above, the attribute gage study involves repeated measuring of the selected parts.

 If you want to use the regression method for testing the hypothesis that bias equals 0, then you must have a minimum of 15 trials. The AIAG method for this test requires exactly 20 trials. The results from these trials can be analyzed as raw data (for example, one case for each trial) or aggregated (one summary case for each part).

Tolerance limits

In Statistica , the tolerance limit is used in calculating bias (see Computational Details).

Although you can specify both an upper and lower tolerance limit, only one limit is needed for the calculations. In general, a lower tolerance limit is used when the part with the lowest reference value has 0 acceptances and the part with the greatest reference value is accepted in every trial. An upper tolerance limit is specified when the opposite is true (for example, the part with the lowest reference value is always accepted and the part with the highest reference value is never accepted).