Appendix 13: SPCHART
SPSS creates nine types of Shewhart control charts. In this appendix, the charts are
grouped into five sections:
• X-Bar and R Charts
• X-Bar and s Charts
• Individual and Moving Range Charts
• p and np Charts
• u and c Charts
For each type of control chart, the process, the center line, and the control limits
(upper and lower) are described.
Notation
The following notation is used throughout this appendix unless otherwise stated:
σ Population standard deviation for measurements X
A Number of sigmas specified by the user, 0 ≤ A ≤ 9
K Number of subgroups
ni Number of units (samples) for subgroup i
N Total sample size, equal to n1 +K+ n K
xij Measurement (observation) for the jth unit (sample) of subgroup i
ni
Mean of measurements for subgroup i, x = ∑ x / n
xi
i
i =1
ij i
Si ni
Sample standard deviation for subgroup i, S = ∑ 3 x − x 8 / 1n − 16
2
2
i ij i i
i =1
Ri Sample range for subgroup i, R = max4 x , K, x 9 − min4 x ,K, x 9
i i1 ini i1 ini
LCL Lower Control Limit
UCL Upper Control Limit
1
2 Appendix 13
Weight
Weights can be used when the data organization is Cases are units.
• Each value for weight must be a positive integer.
• Cases with either non-positive or fractional weights are dropped.
• When weight is in effect, ni is a weighted sum for all the units in subgroup i
and xi and x are weighted means.
X-Bar and R Charts
When X-Bar and R charts are paired, the sample range statistic R is used to
construct the control limits for the X-Bar chart.
Note: Subgroups whose sample sizes are less than the specified minimum value are
dropped.
Equal Sample Sizes
Assume that ni = n for i = 1,K, K . The process for the X-Bar chart is
; @
xi : i = 1,K, K . The center line for an X-Bar chart is the grand mean statistic x :
∑x
1
x= i
K
i =1
and the control limits are
3 0 5 n8
LCL = x − AR / d2 n
UCL = x + AR / 3 d2 0n5 n 8
Appendix 13 3
where
∑R
1
R= i
K
i =1
is the mean range statistic. The process for an R chart is Ri : i = 1,K, K . The; @
center line for an R chart is R and the control limits are
310 5 0 56 8
LCL = max R 1 − Ad3 n / d2 n , 0
UCL = R 11 + Ad 0n5 / d 0n56
3 2
The auxiliary functions are
05
d2 n =
I4 ∞
−∞
1
1− 1− Φ x 0 56 − 1Φ0 x 56 9 dx
n n
II 1 − 1Φ0 x 56 − 21 − Φ0 y57 + 2Φ0 x5 − Φ0 y57 dydx − 1d 0n56
1
d 0 n5 = 2
∞ x n n n 2
2
3
−∞ −∞
2
05
Φz =
I−∞
x 1 −u2 / 2
2π
e du
Unequal Sample Sizes
The processes for X-Bar and R charts are the same as described in the section
“Equal Sample Sizes” above. The center line for an X-Bar chart is the grand mean
statistic x (numerically identical to that in the section “Equal Sample Sizes”):
∑n x
1
x= i i (1)
N
i =1
4 Appendix 13
and the control limits for subgroup i are
LCL = x − Aσ$ / ni
UCL = x + Aσ$ / ni
1 6
The center line for an R chart for subgroup i is Ri = σ$ d2 ni for i = 1,K, K where
∑ 2 R / d 1n 67
1
σ$ = i 2 i
K
i =1
and the control limits for subgroup i are
2
LCL = max Ri − Aσ$ d3 ni ,0 1 67
UCL = Ri + Aσ$ d3 ni 1 6
X-Bar and s Charts
When X-Bar and s charts are paired, the sample sandard deviation is used to
construct the control limits for the X-Bar chart.
Equal Sample Sizes
; @
Assume ni = n . The process for the X-Bar chart is xi : i = 1,K, K . The center line
for an X-Bar chart is x and the control limits are
3 0 5 n8
LCL = x − AS / c4 n
UCL = x + AS / 3c4 0n5 n 8
Appendix 13 5
; @
The process for an s chart is Si : i = 1,K, K . The center line for an s chart is
∑S
1
S= i
K
i =1
and the control limits are
4 1 0 56 9 / c 0n5 , 0
LCL = max S 1 − A 1 − c4 n
2
4
41 − 1c 0n56 9 / c 0n5
2
UCL = S 1 + A 4 4
The auxiliary function is
05 2 Γ n/2 0 5
c4 n =
10 5 6
n −1 Γ n −1 / 2
where Γ(.) is the complete Gamma function.
Note: When n ≥ 25, c4 n05 n can be approximated by n − 0.5 ,
41 − 1c 0n56 9 / c 0n5 can be approximated by 1 /
4
2
4 05
2 n − 2.5 , and c4 n can be
approximated by 04n − 55 / 0 4n − 35 .
Unequal Sample Sizes
The processes for X-Bar and s charts are the same as the processes in the section
“Equal Sample Sizes” above. The center line for an X-Bar chart is x (as defined in
equation (1)) and the control limits are
LCL = x − Aσ$ / ni
UCL = x + Aσ$ / ni
or
3 1 6 ni 8
LCL = x − ASi / c4 ni
UCL = x + ASi / 3c4 1ni 6 ni 8
6 Appendix 13
where
∑ S / c 1n 6
1
σ$ = i 4 i
K
i =1
However, the center line for an s chart for subgroup i is Si = σ$ c4 ni 1 6 for
i = 1, K, K and the control limits are
1 − 2c4 1ni 672 , 0
LCL = max Si − Aσ$
UCL = Si + Aσ$ 1 − 2c4 1ni 672
or
1 − 2c4 1ni 672 / c4 1ni 6, 0
LCL = max Si − ASi
UCL = Si + ASi 1 − 2c4 1ni 672 / c4 1ni 6
Individual and Moving Range Charts
When a weight variable is specified, each unit of the process is expanded to
multiple units based on the case weight associated with this particular unit. The
span (specified by the user) is associated with the expanded process. If the span is
greater than N (the total number of units of the expanded process), an error message
is displayed and neither an Individual nor a Moving Range chart is generated.
Since each subgroup has only one unit, the process for an Individual chart is
; @
yi : i = 1,K, N where yi is the ith unit of the expanded process. For a span of
length m, the moving ranges, are
Ri =
%&max1 y i − m +1 , K, yi 6 − min1 y i − m +1 , K , yi 6 if i = m,K, N
'SYSMIS if i = 1, K, m − 1
Appendix 13 7
The average moving range is
5∑R
1
R=
0 N − m +1
m
i
The center line for an Individual chart is x and the control limits for an Individual
chart are
05
LCL = x − AR / d2 m
UCL = x + AR / d 0m 5 2
The process for a moving range chart is {Ri, i = m,..., N}. The center line for a
moving range chart is R . The control limits for a moving range chart are
31 0 5 0 56 8
LCL = max R 1 − Ad3 m / d2 m , 0
UCL = R 11 + Ad 0 m5 / d 0 m56
3 2
p and np Charts
The data for p and np charts are attribute data. Each measurement xij is either 0 or
1, where 1 indicates a non-conforming measurement. Therefore,
ni
xi + = ∑x ij
j =1
is the count of non-conforming units for subgroup i. When a weight variable is
specified, xi+ is a weighted sum of non-conforming units. If the data are
aggregated and the value of the count variable is greater than the total number of
units for any subgroup, this subgroup is dropped.
8 Appendix 13
Equal Sample Sizes
; @
Assume ni = n The process for a p chart is pi : i = 1,K, K where pi = xi + / n . The
center line for a p chart is
∑p
1
p= i
K
i =1
and the control limits are
LCL = max p − A 2 p01 − p57 / n , 0
UCL = min p + A 2 p01 − p 57 / n , 1
; @
The process for an np chart is xi+ : i = 1,K, K . The center line for an np chart is
∑x
1
x= i+
K
i =1
and the control limits are
4 0 5 9
LCL = max x − A np 1 − p , 0
UCL = min4 x + A np01 − p 5 , n9
Unequal Sample Sizes
; @
The process for a p chart is pi : i = 1,K, K where pi = xi + / ni . The center line for
a p chart is
K K
∑ ∑n p
1 1
p= xi + = i i
N N
i =1 i =1
Appendix 13 9
and the control limits for subgroup i are
2 p01 − p57 / n , 0
LCL = max p − A i
UCL = min p + A 2 p01 − p 57 / n , 1
i
; @
The process for an np chart is xi+ : i = 1,K, K . However, the center line for an np
chart for subgroup i is ni p . The control limits for subgroup i are
LCL = max ni p − A 2n p01 − p57 , 0
i
UCL = min n p + A 2n p01 − p 57 , n
i i i
Note: A warning message is issued when an np chart is requested for subgroups of
unequal sample sizes.
u and c Charts
Measurements xij show the number of defects for the jth unit for subgroup i.
Hence,
ni
xi + = ∑x ij
j =1
is the total number of defects for subgroup i. When a weight variable is used, xi+ is
a weighted sum of defects.
10 Appendix 13
Equal Sample Sizes
;
Assume ni = n . The process for a u chart is ui : i = 1,K, K @ where u i = xi + / n .
The center line for a u chart is
∑u
1
u= i
K
i =1
and the control limits are
3
LCL = max u − A u / n , 0 8
UCL = u + A u / n
; @
The process for a c chart is xi+ : i = 1,K, K . The center line for a c chart is
∑x
1
c= i+
K
i =1
and the control limits for a c chart are
3
LCL = max c − A c , 0 8
UCL = c + A c
Appendix 13 11
Unequal Sample Sizes
; @
The process for a u chart is ui : i = 1,K, K where ui = xi + / ni ui = xi+ / ni. The
center line for a u chart is
∑x
1
u= i+
N
i =1
and the control limits are
3
LCL = max u − A u / ni , 0 8
UCL = u + A u / ni
; @
The process for a c chart is xi+ : i = 1,K, K . The center line for subgroup i is ni u
and the control limits are
3
LCL = max ni u − A ni u , 0 8
UCL = ni u + A ni u
Note: A warning message is issued when a c chart is requested for subgroups of
unequal sample sizes.
References
ASTM Special Technical Publication (STP) 15D 1990. Manual on presentation of
data and control chart analysis. 6th ed. Prepared by Committee E-11 on Quality
and Statistics, American Society for Testing and Materials. Philadelphia, Penn.
Grant, E. L., and Leavenworth, R. S. 1980. Statistical quality control: 5th ed. New
York: McGraw-Hill.
Harter, H. L. 1969. Order statistics and their use in testing and estimation,
Volumes 1 and 2. Aerospace Research Laboratories, United States Air Force.
Washington, D.C.: U.S. Government Printing Office.
