How Do You Do a Capability Study with Few Samples?
- This topic has 1 reply, 1 voice, and was last updated 8 years, 1 month ago by Joel.
- May 22, 2012 at 12:52 pm #54070
dan weigelParticipant@weigel Include @weigel in your post and this person will
be notified via email.
I have 9 parts to study our press height capability. These are rectangles that will be machined later. We are trying to reduce blank size. As you can see there are very few readings – 9 if I count individual parts 18 if I measure far left and far right. Can I perform a capability study with so few data points?0May 22, 2012 at 2:13 pm #193376
You can perform a capability study, but likely not a particularly accurate one. Do you have historical data on this measurement to suggest a distribution? If not, I wouldn’t bother fitting one as you’ll probably get a decent fit on multiple distributions. I’d go with your gut in that case…if normal seems reasonable based on your small sample, use normal. Likewise for weibull (like a normal but usually skewed) or exponential (highly skewed).
Most important, when you do the capability analysis, choose to get confidence intervals on the capability statistics. If you can’t get good estimates, the least you can do is at least quanitify how uncertain you are of them.
I’ll avoid tangential topics like “Did you perform MSA?”, “Why can’t you get more parts?”, etc. and just read things at face value.0
You must be logged in to reply to this topic.