Samples

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Samples
Message posted by RWB on August 14, 2000 at 12:00 AM (ET)

What are the minimum number
of samples required to obtain
reliable control limits?

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Samples
Message posted by Phil on August 16, 2000 at 12:00 AM (ET)

Most SPC textbooks seem to like 20 or more samples of size 4 or 5. That means about 80 to 100 observations.

From a statistical theory point of view, the control limits are supposed to represent a process that is "in-control" and therefore only experiencing common cause (random) variation. It should then be Normally distributed with mean = mu, and standard deviation = sigma. It makes sense that you would want a random sample of more than 50, but closer to a hundred, items from the population to get a good estimate for mu and sigma (the parameters of the Normal Distribution).
(In my experience with sampling from production processes, I found that samples of 50 didn't make good histograms. Samples of 100 or more make better histograms and begin to show normality better).

Other opinions?

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.