## Sample Sizes

## Sample Sizes

(OP)

The QA guys at my company (Automotive Supplier), are always running samples and doing studies to evaluate all sorts of things. They've stated a 32 piece sample is a statistically significant size and they use that size all the time. In my memory from Engineering School, the sample size had to be taken with respect to the population, and the confidence interval. Can anyone shed some light on sample sizes? Any good references for applied quality techniques in a manufacturing enivronment. I appreciate your thoughts.

## RE: Sample Sizes

If you Customer does not give you a sample size, the minimum size for a normal distribution is 30 but I would suggest 50.

Here is the problem though. Some companies think that they need 30 pieces per cavity and so some plastic manufacturers with 156 cavities in a mould sure have a big job.

Make sure that the number is homongenous. In other words, If you have a machine that makes 2 parts each cycle, your sample size is an even number, say 50. If your equipment makes 3 parts per cycle, then your sample size may be 60 - 20 from each station.

When reviewing your histogram or capability study, make sure that you have a bell curve (normal). If 1 station is out to lunch, I would suggest that you ask the tooling people to change the average to match the other group otherwise your Pp and Ppks would be rather small.

Remember, your Pp and Ppk must have a minimum value of 1.67 or 5 estimated standard deviations.

Hope this helps and good luck!!

Dave D.

www.qmsi.ca

## RE: Sample Sizes

The class taught a simplified method for run charts and control limits that had only taking averages and multiplying by constants from tables. They took all the mystery out. Google QualPro.

## RE: Sample Sizes

To develop your estimate of the standard deviation, one uses n-1 and it is somewhat significant with 30 pieces from using n. Once a sample size is over 50 the n-1 (for estimating the standard deviation) and n (standard deviation for the population) is pretty well the same which is the reason why having an extremely large sample size is a waste of time.

Just some trivial information.

Dave D.

www.qmsi.ca