Capability Studies - Do I need to use subgroups?



How should the sample size for a capability study be determined? Do I need to use subgroups, or it sufficient to just measure 30 parts? Can I get any usable data about a process capability by measuring each part in a short run of say 50 parts?

Thanks in advance for your help.

Rick Goodson

If you are doing this for an automotive customer, you should use the AIAG SPC book for guidance. It will save you a lot of problems dealing with your customer. Never the less, on to your question.

You do need to use subgroups. The reason is to long and complicated to explain here. If you want further info get a hold of Grant & Leavenworth's Statistical Quality Control, ISBN 0-07-844354-7 and read chapters three and four on why control charts work.

The subgroup size should be at least 4. At a size of 4 and even with a universe (read process) that is far from normal, the distribution of the average of the samples will be very close to normal. The number of subgroups should be at least 25. Again, there is a statisical basis for this number in Grant and Leavenworth. The frequency between subgroups needs to be sufficient for variation between subgroups based on common cause to take place. Rule of thumb. Spread the subgroups out over 1000 pieces or over one shift, which ever is fewer or smaller.

Hope this helps.
Top Bottom