Standard "bogey" parts for an attribute study with 50 parts / 3 people / 3 times

[ozziegood]

Can anyone tell me the standard "bogey" parts for an attribute study with 50 parts / 3 people / 3 times? I can't seem to find it in the AIAG manual. Thanks!

 

[Miner]

What do you mean by "bogey" parts?

You do want a good mix of both good and bad parts, and these parts should represent just either side of the boundary limit for good and bad. Do not use parts that are blatantly bad and obviously good.

 

[fuzzy]

Memory is such a fragile thing...

My bad memory says that I used to spout: 1 good with 9 bad or the inverse, 9 good with 1 bad. The point I thought was to prove that the resolution of the gage is adequate to determine good from bad (even just once). I could be quite behind the times on this, like circa 1998 behind. Anyone else??

 

[Miner]

My bad memory says that I used to spout: 1 good with 9 bad or the inverse, 9 good with 1 bad. The point I thought was to prove that the resolution of the gage is adequate to determine good from bad (even just once). I could be quite behind the times on this, like circa 1998 behind. Anyone else??

I would stay away from any predictable pattern that the operators could detect. 1 bad or 1 good or 50/50 bad/good are predictable and could influence the operators judgement once they detect a pattern.

I recommend changing the ratio randomly for each study.

 

[RESET]

I generally tend to assure one bad part per item studied. Since the parts manufactured by my company are generally "repositioned" ie. made good before shipment, even knowing that one of the 50 parts is bad throws the operator off so that there is no bias. I have never heard that there should be a standard number or percentage of the samples either in or out of spec.