iSixSigma

Gauge R R in transaction services

Six Sigma – iSixSigma Forums Old Forums General Gauge R R in transaction services

Viewing 30 posts - 1 through 30 (of 30 total)
  • Author
    Posts
  • #37786

    Kuttu
    Participant

    Process :
     
    Loan statement is processed by executive A and reviewed by reviewer B. B marks defects in the loan statement(hard copy) and sends back to A for rework. The reviewer counts no of defects (pre-defined) and enters into an on-line data base regularly. In a month 60 loans are processed.Average no of defects per statement is 35 .
     
    Questions on G R&R:
     
    (i)Can we use no. of defects as continuous data for the purpose of GRR?
    (ii)Can we compare no. of defects marked in hard copy vs no. of defects entered in data base with any tool in six sigma. Here the issue is to ensure correctness of on-line data as future analysis will be done based on this data. Can we assume data of hard copy and data of on-line database as data of two different operators, though  both has been done by the same reviewer.
     
    On the other hand if  we take two different reviewers to review the same statement and conduct GRR, we may  get difference in operator to operator performance as more the experience better is the review (should be). Finally we will come up with a  recommendation  that there should be a common understanding of defects and review mechanism among all reviewers. To do that a training-cum-discussion for reviewers should be done. But  this is not an issue at all. Every reviewer is a 100% good reviewer. This assumption is based on 100% quality feedback by customers in last one year for all reviewers. Do you suggest to go with the second approach? Then how to ensure correct data is entered in database?Kuttu ([email protected]

    0
    #112076

    Kuttu
    Participant

    Re-post

    0
    #112167

    Kuttu
    Participant

    Hi great guys!
    Why all of you are keeping quite? Either the question is useless or all of you have no clue how to tackle the issue.
     

    0
    #112173

    Mike R.
    Participant

    Hi Kuttu,
    You have pretty high expectations for forum turn around time.  I’ll take a look today.  Average cycle time for posts on personal attacks is about 6 hours, mode of 15-30 minutes, sigma of 6.  Did we identify the proper distrbution approximation for cycle time data this week, or did that get lost after we wandered into the takt/cycle hypothesis test debate?  It’s not Normal, sure.  Chi-square is too highly negatively skewed.  Poisson most likely.  Please chime in if tht’s not EXACTLY correct everyone.
    Response time for actual Six Sigma questions has been running much longer than that, especially for calculations and applicability of tests.  Stan has mentioned that it is extremely difficult to choose the right test (or at least, he sees many instances where others make mistakes).  If he really knows where people make mistakes, he probably could help you with the right answer.
    FYI,
    The forum has been a little edgy lately.   Though the question is not all that difficult, the level of exactness in any response on this forum has been raised far beyond the level necessary to help you.
    I’ll write you later if no one else jumps in.
    Regards,
    Mike R.

    0
    #112174

    Johnny Guilherme
    Participant

    Mike
    Give us the answer. Maybe because of the “edgyness” the respective people are not posting. Maybe they are just busy working. I keep wondering about the edgyness-were you guys at school together or what???
    Johnny

    0
    #112175

    Mikel
    Member

    What you want to do is an Attribute R&R. Your concerns about experience shouldn’t come into it. A person is on the job and assumed to be trained. Set up 20 – 30 documents that represent what the reviewer would regularly look at. Have all persons who make this judgement participate and have them look at the documents in a random order. All reviewers look at all documents twice.

    0
    #112176

    Mikel
    Member

    Good morning Billy “miker” Graham,
    I am glad to see you are back to yourself today.
    Excellent advice given to the poster! Keep up the good work.
    Hint – you need to learn the meaning of negative skew. You had a 50/50 chance and you did not pass the test. Please try again and we will be glad to give you hints if you need them.

    0
    #112177

    Kutta
    Participant

    Stan,
    After that what will happen? You would suggest to train these guys to reduce variation in review output with the help of checklist kind of thing etc.
    The other part of the question…How to ensure sanctity of data in the system.How to ensure that the reviewer would enter correct data into the online system? Any giudance?Thanks for your reply.
    Appreciate me that finally I got you great guys on discussion forum!!
    Kutta

    0
    #112178

    Mike R.
    Participant

    A gentle correction… wow. Thanks.  I won’t be expecting another, but thanks just the same.
    Sad truth:  I am not Billy, nor 6surfer.  An even cursory look at the style, grammar, inflection, tone, length, etc. would verify this.
    Bill and Surf Dawg and Max, et. al.:  Thank you for your support.  However, I did not join this forum to evangelize, though I have on occasion when I saw fit.  Any perceived need for more of that that on my part is long gone.  However, using the de facto rules of the forum, I must point out:
    1.  Commentary on a subject you know little about should be reserved until you have done sufficient research to add meaningful insight.  Your comments on my execution of my ethical system did not pass that test.  I did not escalate by response.  I will also not respond to further coments on my ethical system or morality: the degree of execution of an ethical system.
    2.  All data and models are wrong.  Some are useful, to paraphrase(Cox?).  The level of accuracy in the execution of most Six Sigma tools is far less than the standards of this forum.  We have texts.  Outright fallicies should be vigorously refuted.  Isn’t there some level of perfection that is unnecessary?  I think so.
    3. Is “crxp” with the a inserted really worse than “crxp”.  Similarly “bxtthead”.  Please misspell the profanity I am accused of uttering so I don’t run afoul again.
    Regards,
    Mike
     

    0
    #112180

    rajanna
    Participant

    Kuttappa?

    0
    #112185

    Darth
    Participant

    Kuttu,
    I agree that this is probably better suited for an attribute Gage R&R study.  I don’t know if you know how to do a Gage R&R so here is a little refresher.  But first a few questions:
    1.  Are you trying to “test” the Executive or the Reviewer or both?
    2.  Do you really only have one Executive and one Reviewer?
    In an attribute Gage R&R you are trying to basically check three things; accuracy of the test subjects against a master or expert opinion, repeatability which the personal consistency of the subject and reproducibility which is the consistency across two or more subjects.
    Typically, you will select 20-30 loan statements for your test.  You will have an “Expert” determine the true number of defects in those statements.  That will serve as your baseline for accuracy.  You then typically select 3 subjects, in this case the Reviewers.  Have each subject review the statement and identify the defects.  Randomize the statements and have them look at them again.  Match up the two trials for later analysis.
    You can now calculate what percent of the time does each reviewer match the expert on BOTH trials.  That is a measure of their accuracy.  You can calculate the percent of time the reviewers match themselves across both trials.  This is the measure of repeatability.  You can calculate the percent of time that all three of the reviewers agree.  This is a measure of reproducibility.  You can calculate what percent of the time all three agree with each other and with the expert.  That will give you an overall measure of accuracy.
    Whether training is an issue or not will be a function of what you calculate.  Yes, training might be a help.  Better operational definitions of defects might help.  Redesign of the form might help.  Visual acuity problems might be a problem.  If your percentages are in the 80-90ish range then you might be OK given it is a human system.
    You also have the challenge of why the Executives are making those mistakes.
    Finally, you can also investigate the accuracy to which the information is being entered into the online data base.  So, you have, what looks like, three separate projects.  Hope this helps.  Good luck.

    0
    #112251

    Johnny Guilherme
    Participant

    Darth
    Where can one get more info on Gage R&R. I must admit I read your posting but it did not mean much to myself. I looked on the blue next to this discussion forum. Unfortunately I could not get in. i.e. what doe sit mean and where can one apply it.
    Johnny

    0
    #112252

    Kutta
    Participant

    Darth,
    That’s a great reply!It is going to help us in a real manner!! Thanks for your valuable time.
    We have many executives but one reviewer for them.We want to test the accuracy of data entered into the system by the reviewer.
    Can I use G RR(Continuous data) for this situation, as we want to analyze R & R based on no. of defects not on no. of defectives where we use A R & R(my understanding).For example every Unit (statement)  has on an average 35 defects.
    Is there any tool to evaluate correctness of data entered into the on-line system? If not can we presume data of hard copy as one set of data and  data of system as the other set of data as both types of data are entered by the same person.
    Kutta

    0
    #112257

    Dog Sxxt
    Participant

    It shall be reproducibility in your case for between systems.

    0
    #112271

    Dog Sxxt
    Participant

    Expecting some spoon-feeded and free servcices for your own issue? Let forget it.

    0
    #112277

    Darth
    Participant

    From your previous post I understood you had two problems.  One, is the reviewer capturing all the defects and two are they accurately entering it into the system.  If there is only reviewer, you can only measure accuracy and repeatability since there is no second reviewer to determine reproducibility.  Before worrying about accuracy into the system, I would worry about accuracy and repeatability of identifying the errors.  No sense worrying about what goes in until you determine whether it is accurate.
    Conceiveably you could set up an experiment whereby you have one column of the experts’ determination of number of defects for each of 30 documents.  Columns 2 and 3 would be the Reviewers count of the documents…two trials with the 30 randomized.  I tried it with the Attribute R&R and it worked fine.  Of course, all you get is accuracy and repeatability.
    I don’t know of any secret way to measure the accuracy of what is put into the system by the Reviewer other than an audit of what they did.  The concern there is also the measurement system.  Is the auditor’s measurement accuracy and repeatability any better than the Reviewer’s?  I would keep this simple and just go for gross magnitudes.  Do the accuracy and repeatability for the Reviewer in spotting errors and then get some gross magnitude of accuracy of data entry via a simple audit.  If the numbers are high or low, make a decision whether you have a problem and don’t worry so much about MSA at this point.  The measurement system is not that complex in your case.

    0
    #112305

    Dog Sxxt
    Participant

    You may want to review your undersranding on reproducilibity concept in MSA. A system can be consist of reviewer(man) or instrument(machine), etc. 

    0
    #112306

    Dog Sxxt
    Participant

    A system means a measurement system.

    0
    #112310

    Darth
    Participant

    Don’t you need two similar things to test for reproducibility?  If there is one Reviewer and one keyboard, I don’t see how you will test for reproducibility with an MSA study.  Maybe you can enlighten.

    0
    #112311

    Dog Sxxt
    Participant

    My case example is for manual recording and data entry into computer. One reviewer, two data entry apparatus. Hope we talk to a same thing.
     

    0
    #112313

    Kutta
    Participant

    Great Guys,
    Kudos!
    You have been successful in confusing a real issue faced in this kind of scenario.
    Can anyone sum up what exactly I should do?
    -Kutta

    0
    #112318

    Dog Sxxt
    Participant

    Pardon me, my response to Darth is neither to help or confuse you.

    0
    #112323

    mjones
    Participant

    Kutta/Johnny G-
    Darth gave you a very good response on Dec 10.
    Johnny G- To find more info, start with the dictionary on this site under Gage R&R, read the related posts and their references. Also, if you have Minitab, dig into “help” as a great resource. There are examples and explanations there. To really learn this subject, try and take a class or read some books and do some actual applications — just reading posts can be a slow and confusing way to learn (as is evident as you try and understand the Darth/DogS exchanges… and they really do know what they are saying to each other).
    Kutta- By definition, an Attribute Gage R&R is NOT for continuous data. There may be some measures of “data entry time” that give you continuous data, but the measurement of time is not the issue that you describe, rather, the definition of a product defects. So you are back to an Att Ga R&R. The first and obvious requirement (as mentioned in prior posts) is to have a clear, specific, consistent, operational definition of “defect.” Then, you can run a Ga R&R that could be useful. Without the solid definition, expect results to tell you that decisions are inconsistent.
    Good luck.

    0
    #112324

    Darth
    Participant

    If you are still confused and want to take this offline, please feel free to contact me at [email protected] for further clarification.
    The reason you get seemingly confusing responses is that the issue has not been clearly laid out so respondents will be answering differently based on their interpretation of the facts.

    0
    #112325

    Mikel
    Member

    mjones,
    Your advice is generally good, but I would not recommend definitions posted on this site. Two things – 1) there is no good advice on MSA there, and 2) there are many wrong or misleading definitions. While I think Mike is trying to provide a valuable service, he would do well to enagage several of us to clean up his references and definitions.
    The AIAG books and the help menus in Minitab are excellent.
    The main thing most of these folks need to be told about MSA is that it is about the System, not just two operators. This means that ultimately that all equipment, all operators, all judgements need to be quantified. It is the point that Dog Sxxt was making to Darth yesterday.

    0
    #112327

    Mikel
    Member

    Dear DrDarth66,
    I am still confused.
    Can you explain the difference between Measurement Systems Analysis, Variable R & R, and Attribute R & R?

    0
    #112332

    Darth
    Participant

    Mr. Stans,
    If you don’t know by now, then it would be a waste of my time to tell you.  Do you think you can so easily draw me into your little spider’s web?  Do you think by dangling that tasty little bait, I will bite at your hook?  Do you think I am destined to become the son of Mike R?  Do you think the use of the DrDarth66 title will get me to lower my defenses?  Do you think filling my glass with Grappa will get me to open my mouth?  Ok, now that we have settled that.
    I think we all agree that MSA is intended to cover the entire Measurement System, hence its name.  It should also include accuracy, repeatability, reproducibility, stability and linearity.  By its name, it is all inclusive.  The Gage R&R studies are more focused, specifically repeatability and reproducibility, hence the R&R.  The Variable also allows for some interaction effects as well, for example between operators and parts.  Your previous point clarified what Dog Sxxt was saying about worrying about the entire system not just the Reviewer.  I thought I made that point in my posts.  Thanks for the clarification.  Now go hunt for some easier prey to pounce on.

    0
    #112335

    mjones
    Participant

    Stan-
    I absolutely agree with you. Thanks for the emphasis on some critical points…
    A poster was saying he/she could not find anything on the site, and I was saying, look again, some info is here; use it as a starting point, but you need go further… I don’t think this site was designed to be comprehensive. And, yes, maybe we can help Mike with this, add some papers, etc., but much of the learning will come from actually doing things.
    Your other point is more important: it is the system, not just the gages/methods/people, that must be evaluated and improved.
    (I think the initial post proposed doing analysis of the current situation without even defining criteria for ‘good/bad’ products, did not understand how to set up data to do a valid Ga R&R, much less appreciate issues related to a solid MSA. The apparent lack of understanding of the Darth-DogS exchange regarding the critical pieces of the system further suggests needs to further study and understand MSA vs. basic Ga R&R.)
    Regards!

    0
    #112336

    Mikel
    Member

    The answer to the first paragraph is yes I think (or hope) I can.
    The answer to the second paragraph is I just really want someone to come out and play. Please will you come play?

    0
    #112337

    Mikel
    Member

    The running of an attribute study before defining good/bad is always fun and I highly recommend it to those who treat their own operations the same way. Another fun exercise in this type environment is to run the attribute study with the process experts first. I am willing to take bets on the results sight unseen.
    I took the dialogue between Dog and Darth as Dog was using Darth as a surrogate for Mike Carnell since he generally likes to argue with Mike. That bum Mike has been working too much and now has the audacity to go play in Mexico – what’s Dog to do?

    0
Viewing 30 posts - 1 through 30 (of 30 total)

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