Você está na página 1de 6

Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?

ID=638 Page 1


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs


Definitions

LSL = lower specification limit
USL = upper specification limit
x = estimate of process mean
Within
o = StDev(Within) = estimate of within subgroup process standard deviation
n = total number of (nonmissing) observations
n
i
= number of (nonmissing) observations in subgroup i
c
4
(n
i
) = unbiasing constant for subgroups of size n
i
(for use with sample standard
deviations)
Overall
o = StDev(Overall) = estimate of overall process standard deviation
P(X) = probability of event X
Z = standard normal variable
x
i
= the ith observation
x
ij =
the jth observation of the ith subgroup


Note: If the subgroup size is 1,
Within
o is estimated using the average moving range by default. If the subgroup
size is greater than 1,
Within
o is estimated using a pooled standard deviation by default. See Methods of
Estimating Sigma below.


Note: c
4
(n
i
) is an unbiasing constant whether or not the sample size is small or large. The bias is negligible for
large samples (see, e.g. Farnum), as with larger samples, c
4
(n
i
) approaches the value of one.

You can estimate the standard deviation with or without unbiasing constants. By default, Minitab uses
unbiasing constants when estimating the within subgroup standard deviation, and by default does not use
unbiasing constants when estimating the overall standard deviation. To change the default options for the
current session choose Stat > Quality Tools > Capability Analysis > Normal > Estimate. To change the
default options for future sessions, choose Tools > Options > Control Charts and Quality Tools >
Estimating Standard Deviation.








Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?ID=638 Page 2


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs


Formulas

Benchmark Zs

Potential (Within) Capability
Within
LSL x
LSL Z
o
.

= L P C

3 =

Within
x USL
USL Z
o
.

= U P C

3 =

( )
USL LSL
p p Bench Z u =

1 .
1


where:
LSL
p = ) Pr( LSL X < ( ) LSL Z. 1 u =

USL
p = ) Pr( USL X > ( ) USL Z. 1 u =

u(Z) is the cumulative probability
u
-1
(Z) is the inverse cumulative probability

For benchmark Zs for Overall Capability, simply replace
Within
o , L P C

, and U P C

with
Overall
o , L P P

, and
U P P

respectively.

Confidence Intervals for Benchmark Zs

Formulas for Confidence Intervals

Z.bench
( )
L
P LowerBound
1
u = , ( )
U
P UpperBound
1
u =
where:
1
u is the inverse cdf of a standard normal distribution
( )
( )
2 2 1 1 2
2 2 1 1
, ,
, ,
o
o
v v v v
v v v
F
F
P
L
+

=
( )
( )
2 2 1 1 2
2 2 1 1
1 , ,
1 , ,
o
o
v v v v
v v v
+

=
F
F
P
U

( ) 1 2
1
+ = PN v
( ) PN N = 2
2
v
N = total number of observations
P = Prob(observation is outside of the specification limit) based on the normal distribution






Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?ID=638 Page 3


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs


PPM

i) Observed Performance PPM < LSL = 1,000,000 * ((number of observations < LSL) / n)

ii) Observed Performance PPM > USL = 1,000,000 * ((number of observations > USL) / n)

Observed Performance Total = sum of (i) and (ii)


iii) Exp. Within Performance PPM < LSL = 1,000,000 *
( )
|
|
.
|

\
|
>
Within
x LSL
Z P
o


iv) Exp. Within Performance PPM > USL = 1,000,000 *
( )
|
|
.
|

\
|
>
Within
x USL
Z P
o


Exp. Within Performance Total = sum of (iii) and (iv)


v) Exp. Overall Performance PPM < LSL = 1,000,000 *
( )
|
|
.
|

\
|
>
Overall
x LSL
Z P
o


vi) Exp. Overall Performance PPM > USL = 1,000,000 *
( )
|
|
.
|

\
|
>
Overall
x USL
Z P
o


Exp. Overall Performance Total = sum of (v) and (vi)


Note: % figures are obtained by dividing the formulas above by 10,000.


Note: If you use a historical mean, that value will be used in place of X for all PPM (Expected).

If you use a historical sigma, that value will be used in place of
Within
o for all PPM (Expected Within).






Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?ID=638 Page 4


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs


Methods of Estimating o

Overall standard deviation



Estimating o
Within
when subgroup size > 1

Pooled standard deviation Default

degrees of freedom

Note: When the subgroup size is constant, the pooled standard deviation = (square root of the average
of the variances) / c
4
((n
i
-1)+1):



d = n k + 1 degrees of freedom
k = number of subgroups





Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?ID=638 Page 5


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs


Rbar - Average of subgroup ranges

if all n
i
are the same

Sbar - Average of subgroup standard deviations



Note: When the subgroup size is constant, sbar = (average of the subgroup standard deviations) / c
4
(n
i
):



k = number of subgroups





Knowledgebase ID 638: http://www.minitab.com/support/answers/answer.aspx?ID=638 Page 6


Technical Support Document
Capability Analysis (Normal) Formulas Benchmark Zs

Estimating o
Within
when subgroup size = 1

Average moving range Default


MR
i
= the ith moving range = max[x
i
, , x
i-w+1
] min [x
i
, , x
i-w+1
], for i=w, , n



w = the number of observation used in the moving range. w=2 by default.

Median moving range


MR
i
= the ith moving range = max[x
i
, , x
i-w+1
] min [x
i
, , x
i-w+1
], for i=w, , n

= the median of all MR


i

w = the number of observations used in the moving range. w=2 by default.

Square root of MSSD Square root of half of the Mean of the Squared Successive Differences



d
i
= successive differences

c
4
(n
i
)
|
~ c
4
(n
i
)


References:
Montgomery, D.C. (2004). Introduction to Statistical Quality Control, 5
th
Edition. John Wiley & Sons.
Donald J. Wheeler(1995). "Advanced Topics in Statistical Process Control", SPC Press, Inc.
N. R. Farnum (1994). Modern Statistical Quality Control and Improvement, Wadsworth Publishing.
I.W. Burr (1976). Statistical Quality Control Methods, Marcel Dekker, Inc.