Home › Forums › Old Forums › General › Sample size for process capability studies
This topic contains 4 replies, has 3 voices, and was last updated by Luis 12 years, 7 months ago.
I found in some manuals diferent numbers of minimum sample size for process capability studies. How much samples is necessary?
Sample size affects how precisely you estimate the true process capability index (Cpk). There is no hard and fine rule. You can live with 30 samples but take note of your precision.
Sorry, I do not have the formula to compute the precision.
Maybe the experts here can help.
Rgds
30 pieces is usually enough for a capability study…(unless of course, you have attributes data, then it can get quite large, especially if you have a very low defect rate…)
For variables data, my experience has been that the concern about accuracy or usefullness as a predictor of future performance (of which precision is only a part) has more to do how the samples are collected and not the number of samples. If the sample is not RANDOM and therefore representative of the true range and type variation – your cpaability value can be very ‘overstated’ and misleading.
For example, 30 sequential pieces from a new die for a stamping operation is NOT representative as there will be very little variation within the sample. Ditto, taking 30 pieces from one vendor batch of raw material when the largest component of variation is batch to batch will yield a great number that is totally wrong.
You needto understand the largest componenet of variation to ensure your sampling encompasses the full range of variation. If you dont’ know (as with a new or totally undocuemtned process) then your study needs to have informed input from the subject matter experts as to what kind of variation can be expected to take place and it shoudl take place over some time…
And don’t forget that it must come from a stable process…
The book Measuring Process Capability – Davis Bothe shows equations to calculate the minimum required sample size and the lower confidence bound (considering that the process is stable and the distribution is normal).
If I take 30 samples and Cp estimated =1 we found the lower confidence bound = 0,78, is this precision acceptable?
The book Measuring Process Capability – Davis Bothe shows equations to calculate the minimum required sample size and the lower confidence bound (considering that the process is stable and the distribution is normal).
If I take 30 samples and Cp estimated =1 we found the lower confidence bound = 0,78, is this precision acceptable?
The forum ‘General’ is closed to new topics and replies.
© Copyright iSixSigma 2000-2016. User Agreement. Any reproduction or other use of content without the express written consent of iSixSigma is prohibited. More »