Ed ..wow..you have very close to Iam expecting in my mind...
let me give you more input to clean the stats further..
Data collection will look like this:
Answer Key:
Sample Good Scratches Stains Pits Chips
#1 1 0 0 0 0
#2 0 1 0 0 0
#3 0 0 0 1 0
.
.
.
#10 1 0 0 0 0
Operator#1(Given the samples for Trial 1)
Sample Good Scratches Stains Pits Chips
#1 0 0 1 0 0
#2 0 1 0 0 0
#3 1 0 0 0 0
.
.
.
#10 1 0 0 0 0
Simillarly Trial 2 and 2.
So this data of 3 trails help me to identify the waekness of the
operator in identifying a good part and not being able to identify
the exact defects.
If the averages of the 3 trails of good and corresponding defects donot
coincide with Key answer, I conclude that the operator has weakness on
that particular defect(s).
Now I can modify my analysis with confidence level and uncertainty.
What do you think? Please help me to analyse professionally without any
bias.Also please indicate the source of that formula on Uncertainty
"17% / SQRT(2*(N-1))= 2.2%"
Thanks,
Govind.
In article <8uqdqq$558$[EMAIL PROTECTED]>,
Ed Vojcak <[EMAIL PROTECTED]> wrote:
>
>
> You'll find that this is not covered in standard text.
> The closest analogy I could think of is the uncertainty
> (Repeatability) of a point count, similar to error
> encounted on X-ray scintillation counters.
>
> For your case let the 10 samples A - J give:
>
> Sample DATA Error
> A 000 0
> B 111 0
> C 010 1
> D 100 1
> E 110 1
> F 111 0
> G 111 0
> H 000 0
> I 100 1
> J 001 1
>
> Now you got 10*3 =30 counts with 5 errors or 5/30 = 17% uncertainty.
> However this uncertainty has an uncertainty (confidence limit)
> equal to 17% / SQRT(2*(N-1))= 2.2% where N=30 so your
> repeatability is 15 to 19 %.
>
> Notice to decrease your confidence limits your fighting the
> "inverse square root battle" i.e. to half your uncertainty
> you got to quadruple your point count.
>
> This study should be done several times, over time
> and plotted (really an "p-chart") to get an idea
> of stability. Don't total over time since conditions
> between trials change.
>
> Ed Vojcak PE, CQE
>
> (P.S. "point count laws" applicable to votes also;)
>
> In article <8uphvn$e5r$[EMAIL PROTECTED]>,
> Govind <[EMAIL PROTECTED]> wrote:
> > Brian,
> > I saw MSA from QS 9000 already.That example is not applicable to my
> > case.That example is more towards using a attribute go/nogo gage.My
> > case is visual inspection using microscope which does not tell any
> > intensity of Reject.But I can classify defect.
> > The inspector make judgement based on Specification.
> > Let me explain once again clearly..
> > I have answers key with me for the product samples.
> > These samples are given to inspectors to inspect at random -3
trials.
> > If they correctly detect, I mention as "1" if not I mention "0".
> > with this 1,0 colums, how do I translate to R & R?
> > Govind.
> >
> > In article <4bqP5.5179$[EMAIL PROTECTED]>,
> > "Brian S Hoff" <[EMAIL PROTECTED]> wrote:
> > > Hello Mr. Govind,
> > >
> > > I hope I am pointing you in a proper direction with this
suggestion.
> > >
> > > I believe additional information for R&R studies on attribute
> > inspection can
> > > be found in the QS-9000 Measurement System Analysis manual.
> > >
> > > I believe you could purchase the manual through the ASQ, AIAG, or
> one
> > of the
> > > engineering document services that can be found online.
> > >
> > > I am responding from home so I'm afraid I cannot specifically
> address
> > your
> > > question. (The manual is back in my office.)
> > >
> > > Hope this helps. Good luck :)
> > >
> > > --
> > > Brian
> > >
> > > Govind <[EMAIL PROTECTED]> wrote in message
> > > news:8ukt2c$u33$[EMAIL PROTECTED]...
> > > > Dear Experts,
> > > > I want to conduct a R & R study for attribute charaectistics
like
> > > > Scratches, pits,stains, etc..
> > > > Only Microscope and Inspector are involved.
> > > > What I did was, I took 10 different samples with different type
of
> > > > defects made every inspector inspect 3 times each..
> > > > 1-10 and 1-10 and again 1-10.(in random order).
> > > > Now how do I tranform the results to get Reprodcibility
(appraiser
> > > > variation).I dont suspect Repeatability(gage variation) which
is a
> > > > microscope.
> > > > Can some one help me to transform the results to meaningful
> > > > interpretration.
> > > > If the inspector had correctly detected the defect, I value "1"
> > > > if not "0". So the results are basically 1,0.
> > > > Thanks for patiently go thro' this message.
> > > > Govind.
> > > >
> > > >
> > > > Sent via Deja.com http://www.deja.com/
> > > > Before you buy.
> > >
> > >
> >
> > Sent via Deja.com http://www.deja.com/
> > Before you buy.
> >
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
>
Sent via Deja.com http://www.deja.com/
Before you buy.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
