The standard basic study uses 10 parts, 3 appraisers, and 2 trials, giving 10 × 3 × 2 = 60 readings. The 10 parts should span the full range the process actually produces, not 10 parts from the same bin, because the study compares measurement spread against part spread. Mark each part with a hidden number so appraisers cannot recognize it. Use the production gage, the production method, and the appraisers who normally do the job.
Order matters. Each appraiser measures all 10 parts once in random order, then measures them all again in a different random order. Appraisers must not see each other's readings or their own first-trial readings. A recorder who is not an appraiser writes everything down. Breaking any of these rules lets memory and expectation creep into the numbers and makes the gage look better than it is.
Once the 60 readings are in, the arithmetic is averages and ranges. For each appraiser and each part, compute the range between the two trials; average all 30 of those ranges to get the average range, written R-bar. Compute each appraiser's overall average across all 20 of their readings; the largest minus the smallest appraiser average is the appraiser difference. Average each part across all six readings it received; the largest part average minus the smallest is the part range, written Rp.
Standard Procedure
- 1Select 10 parts covering the normal process spread and number them where appraisers cannot see.
- 2Choose 3 regular appraisers and the production gage; appoint a separate recorder.
- 3Run 2 trials per appraiser in randomized order, blind to previous readings.
- 4Compute R-bar from the 30 trial-to-trial ranges, the appraiser difference from the 3 appraiser averages, and Rp from the 10 part averages.
Lesson Summary
- ✓Random order, hidden part numbers, and no peeking are what make the study honest.
- ✓Three simple numbers, R-bar, appraiser difference, and Rp, summarize all 60 readings.