Estimating Specifications
Six Sigma – iSixSigma › Forums › General Forums › General › Estimating Specifications
 This topic has 5 replies, 4 voices, and was last updated 10 years, 8 months ago by gomezadams.

AuthorPosts

March 26, 2012 at 3:07 pm #54005
Is it possible to estimate the range of individual values from two samples? A customer wants us to do specifications but we only have two samples.
Thanks.
0March 26, 2012 at 4:11 pm #192747
Chris SeiderParticipant@cseider Include @cseider in your post and this person will
be notified via email.Hmmm, it is nice that the customer is willing to make THEIR specifications based on YOUR process variation. Would you assume a normal distribution?
The confidence interval for standard deviation is so wide with n=2, you can’t be sure of your real process variation so setting control limits or spec limits would seem to be full hardy.
My 2 cents…or is it 2 observations. :)
0March 27, 2012 at 2:20 pm #192790
gomezadamsMember@spazwhatsup Include @spazwhatsup in your post and this person will
be notified via email.Isnt this an oxymoron? Estimating specifications?
Chris is right on!
You end up with a process you could throw a football field through.0March 27, 2012 at 3:02 pm #192791
Robert ButlerParticipant@rbutler Include @rbutler in your post and this person will
be notified via email.The short answer is yes. You can use the two samples to generate an estimate of the range R = abs(sample 1 – sample 2). Then you can divide this value by 1.128 to get an estimate of the standard deviation and press on from there.
Unless you have some very compelling reason for doing this – extreme costs with respect to sample generation/testing, a 500 pound gorilla threatening you with severe bodily harm unless you provide an answer, a need for a very quick/rough estimate of how things stand(with the understanding that it is indeed a VERY rough estimate), etc. I would recommend that you not use this to generate spec limits for all of the reasons already cited.
If you need the particulars with respect to the magic number 1.128 and the above reasoning check Quality Control and Industrial Statistics – Duncan 4th edition pp.140141
0March 27, 2012 at 7:48 pm #192795
MBBinWIParticipant@MBBinWI Include @MBBinWI in your post and this person will
be notified via email.Bill – Wow! Specs should be related to the functional need(s) of the design. Specs have nothing to do with part variation (other than comparing variation to range to evaluate capability).
0March 28, 2012 at 2:43 am #192799
gomezadamsMember@spazwhatsup Include @spazwhatsup in your post and this person will
be notified via email.What type of spec is it?
We may be able to provide more meaningful feedback once we know.
eg Is it an electrical parametric measurement,mechanical,….?0 
AuthorPosts
You must be logged in to reply to this topic.