iSixSigma

Subgroups in Sampling

Six Sigma – iSixSigma Forums Old Forums General Subgroups in Sampling

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #28319

    Sandra
    Member

    Are there any guidelines or “rules of thumb” regarding subgrouping?  Is it purely arbitrary or cost driven?  Is it better to take one sample per hour for five days or 10 samples, three time a day for five days?  Is there any literature available on this subject?
    Thank you in advance for your assistance.

    0
    #70367

    Utah
    Member

    Subgrouping is used to reduce the variation within the data so that those external things influencing your process are easier to detect (i.e. Reducing the “within” variation so the “between” variation is more pronounced, if you know what that is).  If there is a chance a portion of your data shares commonalities seperate from the rest of the data stream, subgroup.  Likewise with possible differences.  Just look at the structure of the process to determine where to subgroup (e.g. along shifts, operators, changeovers, machines……look where there are  changes in the process.)  Note:  There is no rule as to how many subgroups you need to have…of course more is better, but plotting whatever you have still provides very useful info.  Also, subgroup size is relevant in that increasing it increases the sensitivity of you control chart, while decreasing it does the opposite.  For example, if you had only a few subgroups plotted but n = 50, then anything outside control limits would definetly be deemed “out of control” and reason to investigate for special cause.  Pick the size according to the characterisitics of the data and its costs.  A great reference is SPC by Wheeler…..great info, not to tough to grasp.  Good Luck.

    0
    #70368

    denton
    Participant

    If you subgroup correctly, Xbar and R charts will show you things you won’t easily see otherwise.  If you subgroup incorrectly, they will flat out lie to you.  My experience is that there are more Xbar and R charts done wrong than done right.
    A short rule is that you do your subgrouping so that the variation you don’t care about falls within groups, and the variation you do care about falls between groups.
    Common error:  We have 5 machines, so we take a sample fron each machine every hour to form a subgroup.  You have just reduced the machine to machne variation by 5^.5.  Take 5 consecutive samples from one machine, then 5 from another, etc. and you will see a completely different story.
    For whatever reason (or maybe no reason), 5 is a really common subgroup size…
    Hope this helps some.  Xbar and R and Process Capability are two tools that tend to get really sorely abused.
    Denton Bramwell, Sr. Master Black Belt
    [email protected]
     

    0
    #70371

    Leung
    Participant

    In addition to the Wheeler book, you might want to check out Montgomer’s “Introduction to Statistical Quality Control.”
    You should be familiar with the concept of rational subgrouping if you want to get the greatest use of your control charts.
    Ben

    0
    #70372

    Marc Richardson
    Participant

    Denton,When I took my first SPC class in 1976, the instructor, Brian Dawes, said that the subgroup size of 5 arose from the fact that it’s easy to divide by five (multiply by .2). This goes back to pre-calculator days.
    By the way, neither email address I have for you seems to work. I’d like to contact you directly regarding the project on which we were working.Marc Richardson

    0
    #70387

    Sandra
    Member

    Utah,
    Thank you for your response.  It has provided me with more insight into subgrouping and SPC.  I was researching the reference you provided and discovered that Wheeler has authored several books on SPC.  Could you please reply to this message with the specific title of the book you were referring to?  Thank you for your time and assistance.   Sandra

    0
    #70389

    Utah
    Member

    “Understanding SPC” I believe is the text that I have found most useful by Wheeler.  It is comprehensive and straight forward.  All SPC is is Control Charting and how to construct it, use it, and understand it. To me, this along with DOE is the heart of the 6S tool set.  If you have a good handle on these tools (i.e. read Wheeler and Montgomery), the rest is easy.  Good Luck and let me know if I can be of further assistance.

    0
    #70446

    denton
    Participant

    Marc…
    Sorry about the email problems.  I now have working email at [email protected].  The guy that takes care of our email just went through a couple of rounds of surgery, and the anesthetic leaves him severely wiped out for a few weeks.

    0
    #70448

    chew
    Participant

    Hi, glad you guys are discussing about subgroup in SPC.
    My concern is more on Machine SPC rather that Process SPC.
    In Machine SPC, we want to monitor changes or stability of a machine from time to time. Thus, Standards with known reading are generated and tested every say 4 hours. In this particular case, what is the best subgroup size for the Standards. Since within variations are not so crucial in Machine SPC, can we use One for this subgroup ?
    Another question is Machine SPC may need to use different way of calculating Control Limit. Is conventional method like CL = X +/- AR still valid ?
    When we select subgroup Standards, we did not consider the variation within them. We are tracking the variation between this subgroup to see any changes on the tester performance.
    Thanks for your response. cheers….Chew
     

    0
    #70451

    Shree Phadnis
    Member

    HI,
    IN FACT TO GIVE A CONCEPTUAL UNDERSTANDING TO SPC I WROTE THE ARTICLE BASED ON ACTUAL CASE STUDIES EXPERIENCED IN THE INDUSTRY.THE ARTICLE WILL GIVE YOU CONCEPTUAL UNDERSTANDING REGARDING SUB GROUPS.
     
    Shree Phadnis

    0
Viewing 10 posts - 1 through 10 (of 10 total)

The forum ‘General’ is closed to new topics and replies.