Chapter 5

Control Charts For Variables

We look at control charts for variables (as opposed to attributes).

5.1 Introduction
We look at three types of sets of control charts:

• the x̄ (mean) and the R (range) charts.

• the x̄ (mean) and the S (standard deviation) charts.

• for individual observations and the R (range) charts.

We will use these control charts to determine a variety of things, but mostly to
determine the probability of detecting outliers or not, of determining the capability
of a process and of determining normality.

5.2 Control Charts For x̄ and R

We look at the use of the x̄ (mean) and R (range) charts in quality control processes.
We also calculate process capability, Cp , and two versions of average run length:

1. ARL0 : average run length, if the process actually is in control1

2. ARL1 : average run length, if the process actually is out of control2

Exercise 5.1 (Control Charts For x̄ and R)

1
We want this to be big.
2
We want this to be small.

81
82 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

1. Ice Cream Treat Weight, Raw Data

Ice cream treats are weighed (in ounces), four at a time (n = 4).

sample x1 x2 x3 x4
1 7 9 12 14
2 9 5 5 12
3 7 10 11 7
4 8 9 5 14
5 7 8 11 12
6 12 11 9 10
7 16 9 10 7
8 6 5 11 3
9 8 7 10 14
10 15 17 9 14
11 7 12 13 17
12 5 11 8 12
13 16 10 13 14
14 8 13 9 13
15 11 8 11 16
16 15 11 10 14
17 10 9 11 10
18 16 8 11 12
19 8 7 10 13
20 15 17 10 15

SAS program: att4-5-2-icecream-xrchart

(a) x̄ chart and R chart

From SAS, both the x̄ chart and R chart for the ice cream weight is (choose
one) in / out of statistical control because there are no out-of-control
signals, runs, trends, or cycles.
(b) (LCL, UCL) for x̄ chart and R chart
From the SAS graph the average of all averages is,
x̄¯ = (choose one) 10.45 / 10.55 / 10.65
The upper control limit and the lower control limit for x̄ chart are,
√
x̄¯ ± 3σ̂/ n = (LCL, UCL) =

(choose one)
i. (5.78, 15.32)
ii. (5.68, 14.32)
iii. (6.78, 15.32)
Section 2. Control Charts For x̄ and R (ATTENDANCE 4) 83

Also, from SAS, R̄ = (choose one) 6.6 / 7.6 / 8.6

The upper control limit and the lower control limit for the R chart are,
(LCL, UCL) = (choose one)
i. (5.5, 14.9)
ii. (0, 14.9)
iii. (−14.9, 14.9)
(c) Natural tolerance limits, control limits and specification limits
USL
upper specification limits
(predetermined)

UNTL
upper natural tolerance limit
LCL = 15.32
upper control limit

10.55

UCL = 5.78
lower control limit
LNTL
lower natural tolerance limit samples

LSL
lower specification limits
(predetermined)
average x,
individual x sample n = 4

Figure 5.1 (Natural tolerance limits, control limits and specification

limits)
On the one√ hand, the control limits for the x̄ chart are,
x̄¯ ± 3σ̂/ n = (LCL, UCL) = (5.78, 15.32),
whereas, on the other hand, the natural tolerance limits for an individual
x are
(UNTL, LNTL) = x̄¯ ± 3σ̂ = 10.55 ± 3(3.18154) =
(choose one3 )
i. (1.01, 20.09)
ii. (0.68, 24.32)
iii. (−1.02, 25.32)
and the specification limits are (arbitrarily) predetermined limits set as
“goals” to be achieved by the ice cream treat process in this case.
(d) More process standard deviation
In addition to reading the process standard deviation σ̂ from SAS, we
could also use d2 = 2.059 from Appendix VI, page 761 of the text, and so
determine the process standard deviation to be

σ̂ = R̄/d2 ≈ 6.55/2.059 ≈

(choose one) 3.18 / 4.34 / 5.66

3
The process standard deviation σ̂ can be found on the SAS printed output.
84 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

(e) Process standard deviation versus standard error

True / False
The process standard deviation σ̂ = 3.18 does not equal the standard
deviation of the mean, σ̂x̄ = √σ̂n = 3.18
√ = 1.59
4
(f) Process Capability
The process capability, Cp , is a statistic used to measure if the natural
tolerance limits (not the control limits!) meet the specification limits or
not. For example, if we (arbitrarily) set the specification limits to be
(USL, LSL) = (6, 15)
and, from above, since we determined the natural tolerance limits to be
(UNTL, LNTL) = (1.01, 20.09)
then the process capability is, in this case,
USL − LSL
Ĉp =
UNTL − LNTL
15 − 6
= =
20.09 − 1.01
(choose one) 0.47 / 2.12 / 5.66
which indicates the “natural” process limits are a well outside the “pre-
scribed” specification limits (which is “bad”); in other words, the process
is not able to meet specifications.
(g) More process capability
If we (arbitrarily) set the specification limits to be
(USL, LSL) = (0, 21)
the process capability is
USL−LSL 21−0
Cp = UNTL−LNTL = 20.09−1.01 = (choose one) 0.47 / 1.10 / 5.66
In other words, the process (choose one) is / is not able to meet specifi-
cations.
(h) More process capability
True / False
USL − LSL
Ĉp =
UNTL − LNTL
USL − LSL
=
6σ̂
USL − LSL
=
6(R̄/d2 )
15 − 6
≈
6(6.6/2.059)
≈ 0.47
Section 2. Control Charts For x̄ and R (ATTENDANCE 4) 85

(i) Normal probability plot

The SAS normal probability plot is more or less linear and so indicates the
weights of the ice cream treats (choose one) are / are not more or less
normally distributed.
(j) New samples in process control?
Five new samples from the ice cream process are plotted on the previously
determined control charts above.
sample x1 x2 x3 x4
1 17 19 22 24
2 19 14 16 21
3 16 20 21 17
4 19 18 15 24
5 17 18 21 22
From SAS, it appears the ice cream process has gone out of control and,
more specifically, has shifted (choose one) downwards / upwards al-
though the range of the process still appears to be in control.
(k) Probability in control if there is a shift in mean?
The probability the process remains in control if the mean shifts from 10.55
to 12 ounces, is equal to the probability x̄ falls between the upper control
limit and the lower control limit,

P (in control) = P (LCL ≤ x̄ ≤ UCL)

= P (5.78 ≤ x̄ ≤ 15.32; µ = 12, σ̂x̄ = 1.59) =

(choose one) 0.47 / 0.78 / 0.98

Assuming normality, it is 2nd DISTR normalcdf(5.78, 15.32, 12, 1.59)
(l) More probability in control if there is a shift in mean?
The probability the process remains in control if the mean shifts to 4
ounces, is equal to the probability x̄ falls between the upper control limit
and the lower control limit,

P (in control) = P (LCL ≤ x̄ ≤ UCL)

= P (5.78 ≤ x̄ ≤ 15.32; µ = 4, σ̂x̄ = 1.59) =

(choose one) 0.13 / 0.68 / 0.87

Hint: 2nd DISTR normalcdf(5.78, 15.32, 4, 1.59)
(m) Different ways of saying the same thing
The probability the process remains in control if the mean shifts to 4
ounces, is another way of saying (choose none, one or more)
i. the probability of not detecting a mean shift to 4 ounces.
86 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

ii. the probability of not detecting the first sample average after the shift
out of control, if the mean shifts to 4 ounces.
iii. the probability of a type II error, β, or the probability of accidently
rejecting the alternative hypothesis of a mean shift to 4 ounces.
iv. the probability of not detecting the second sample average after the
shift out of control, if the mean shifts to 4 ounces.
v. the probability of not detecting the third sample average after the shift
out of control, if the mean shifts to 4 ounces.
(n) Probability of not detecting first sample average out of control
The probability not detecting first sample average out of control if the
mean shifts to 4 ounces is
P (5.78 ≤ x̄ ≤ 15.32; µ = 4, σ̂x̄ = 1.59) = (choose one) 0.13 / 0.68 / 0.87
(o) Probability of not detecting third sample average out of control
The probability not detecting third sample average out of control if the
mean shifts to 4 ounces is
P (5.78 ≤ x̄ ≤ 15.32; µ = 4, σ̂x̄ = 1.59)3 = (0.13)3 =
(choose one) 0.002 / 0.068 / 0.87
(p) Probability of detecting mean shift
The probability of detecting a mean shift to 4 ounces is
1 − P (5.78 ≤ x̄ ≤ 15.32; µ = 4, σ̂x̄ = 1.59) =
(choose one) 0.13 / 0.68 / 0.87
2. Ice Cream Treat Weight, Summary Data4
Samples of n = 4 units are taken from an ice cream treat process every hour.
After 25 samples have been collected, we find x̄¯ = 10 ounces and R̄ = 7 ounces.
(a) Control limits
The control limits for x̄ are

x̄¯ ± A2 R̄ = 10 ± 0.729(7) =

(choose one)
i. (4.90, 15.10)
ii. (5.68, 14.32)
iii. (6.78, 15.32)
The (estimated) control limits for R are

(D3 R̄, D4 R̄) = (0(7), 2.282(7)) =

(choose one)
4
Since summary data is given, we cannot use SAS to help us in this question and must use
Appendix VI on page 761 of the text.
Section 2. Control Charts For x̄ and R (ATTENDANCE 4) 87

i. (0, 15.10)
ii. (0, 14.32)
iii. (0, 15.98)
(b) Process standard deviation
The (estimated) process standard deviation is
σ̂ = R̄/d2 ≈ 7/2.059 ≈
(choose one) 2.13 / 3.40 / 4.87
(c) Process capability
If the specification limits are (3, 17), the (estimated) process capability is
USL − LSL
Ĉp =
6σ̂
USL − LSL
=
6(R̄/d2 )
17 − 3
= ≈
6(7/2.059)
(choose one) 0.13 / 0.40 / 0.69
which indicates the “natural” process limits are (choose one) inside /
outside the “prescribed” specification limits; the process is not able to
meet specifications.
(d) Probability of not detecting mean process shift
The probability of not detecting a mean process shift from µ = 10 to
µ = 14 is equal to the probability x̄ falls between the upper control limit
and the lower control limit when µ = 14,
P (in control) = P (LCL ≤ x̄ ≤ UCL)
√
= P (4.90 ≤ x̄ ≤ 15.10; µ = 14, σ̂x̄ ≈ 3.40/ 4) =
(choose one) 0.13 / 0.40 / 0.74 √
2nd DISTR normalcdf(4.9, 15.1, 14, 3.4/ 4)
(e) Review: Average Run Length (ARL)
Since µ ≈ x̄¯ = 10 and σ̂ = 3.4, the probability the process is in control is,
√
P (LCL ≤ x̄ ≤ UCL) = P (4.9 ≤ x̄ ≤ 15.1; µ = 10, σ̂x̄ ≈ 3.40/ 4)
= 0.9973
√
2nd DISTR normalcdf(4.9, 15.1, 10, 3.4/ 4)
and so the ARL0 is
1
ARL0 = =
1 − 0.9973
(choose one) 1.13 / 2.40 / 370.37
88 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

(f) Control limits, using α = 0.03 rather than σ

If α = 0.03, the control limits for x̄ are
µ ¶
3.4
x̄¯ ± z0.03/2 σ̂x̄ = 10 ± 2.17 √
4
(choose one)
i. (4.90, 15.10)
ii. (5.68, 14.32)
iii. (6.31, 13.69)
Hint: z0.03/2 = 2.17, using invNorm(0.985)

5.3 Control Charts For x̄ and S

In this section, we look at control charts5 for x̄ and S

Exercise 5.2 (Control Charts For x̄ and S: Ice Cream Treat Weight)
Ice cream treats are weighed (in ounces), four at a time (n = 4).
sample x1 x2 x3 x4
1 7 9 12 14
2 9 5 5 12
3 7 10 11 7
4 8 9 5 14
5 7 8 11 12
6 12 11 9 10
7 16 9 10 7
8 6 5 11 3
9 8 7 10 14
10 15 17 9 14
11 7 12 13 17
12 5 11 8 12
13 16 10 13 14
14 8 13 9 13
15 11 8 11 16
16 15 11 10 14
17 10 9 11 10
18 16 8 11 12
19 8 7 10 13
20 15 17 10 15
5
Notice, we are looking at a control chart for S, for standard deviation, and not R, for range, as
it was done previously.
Section 3. Control Charts For x̄ and S (ATTENDANCE 4) 89

SAS program: att4-5-3-icecream-xschart

1. x̄ chart and S chart

From SAS, both the x̄ chart and S chart for the ice cream weight is (choose
one) in / out of statistical control because there are no out-of-control signals,
runs, trends, or cycles.

2. (LCL, UCL) for x̄ chart and S chart

From the SAS graph, x̄¯ = (choose one) 10.45 / 10.55 / 10.65
The upper control limit and the lower control limit for x̄ chart are,
√
x̄¯ ± 3σ̂/ n = (LCL, UCL) =

(choose one)

(a) (5.78, 15.32)

(b) (5.86, 15.24)
(c) (6.78, 15.32)

From SAS, S̄ = (choose one) 2.9 / 7.6 / 8.6

The upper control limit and the lower control limit for the S chart are,
(LCL, UCL) = (choose one)

(a) (5.5, 14.9)

(b) (0, 14.9)
(c) (0, 6.5)

3. More (LCL, UCL)

The control limits for the x̄ chart and S charts,
(LCL, UCL) = (5.86, 15.24)
is (choose one) the same as / different from
the control limits for the x̄ chart and R charts,
(LCL, UCL) = (5.78, 15.32)
because the process standard deviation, σ̂ is calculated differently in the two
cases.

4. Process standard deviation

In addition to reading the process standard deviation σ̂ from SAS, we could
also use c4 = 0.913 from Appendix VI, page 761 of the text, and so determine
the process standard deviation to be

σ̂ = S̄/c4 ≈ 2.9/0.9213 ≈

(choose one) 3.15 / 4.34 / 5.66

90 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

5. Process capability
If we (arbitrarily) set the specification limits to be

(USL, LSL) = (0, 21)

and since the natural tolerance limits for an individual x are

(UNTL, LNTL) = x̄¯ ± 3σ̂ = 10.55 ± 3(3.12691) = (1.169, 19.931)

the process capability is

USL−LSL 21−0
Cp = UNTL−LNTL = 19.931−1.169 = (choose one) 0.47 / 1.12 / 5.66
In other words, the process (choose one) is / is not able to meet specifications.

6. More process capability

True / False

USL − LSL
Ĉp =
UNTL − LNTL
USL − LSL
=
6σ̂
USL − LSL
=
6(S̄/c4 )
21 − 0
≈
6(2.9/0.9213)
≈ 1.11

7. Probability in control if there is a shift in mean?

The probability the process remains in control if the mean shifts from 10.55 to
17 ounces, is equal to the probability x̄ falls between the upper control limit
and the lower control limit,

P (in control) = P (LCL ≤ x̄ ≤ UCL)

√
= P (5.86 ≤ x̄ ≤ 15.24; µ = 17, σ̂x̄ ≈ 3.12691/ 4) =

(choose one) 0.13 / 0.78 / 0.98

Assuming normality, it is 2nd DISTR normalcdf(5.86, 15.24, 17, 1.56)

8. Probability of not detecting first sample average out of control

The probability not detecting first sample average out of control if the mean
shifts from 10.55 to 17 ounces is
P (5.86 ≤ x̄ ≤ 15.24; µ = 17, σ̂x̄ = 1.56) = (choose one) 0.13 / 0.68 / 0.87
Section 4. The Shewhart Control Chart For Individual Measurements (ATTENDANCE 4)91

5.4 The Shewhart Control Chart For Individual

Measurements
In this section, we look at Shewhart control charts for individual measurements.

Exercise 5.3 (Shewhart control charts for individual measurements)

Thirty (30) individual ice cream treats are weighed (in ounces).
5.4 8.6 9.5 4.0 0.9 5.2 5.5
2.8 5.3 6.3 3.9 9.8 6.9 9.8
5.1 8.4 1.0 1.2 7.1 5.7 0.6
1.0 3.0 6.0 7.2 8.0 5.9 0.0
7.9 3.4
SAS program: att4-5-4-icecream-irchart
1. Individual measurement chart and moving range MR chart
From SAS, both the individual chart and moving range MR chart for the ice
cream weight is (choose one) in / out of statistical control because there are
no out-of-control signals, runs, trends, or cycles.

2. Moving range, MRi

Since the weight of the first ice cream treat is x0 = 5.4 and the weight of the
second ice cream treat is x1 = 8.6, the first moving range (based on n = 2
observations) is
MR1 = |x1 − x0 | = |8.6 − 5.4| = 3.2
and so the second moving range is

MR2 = |x2 − x1 | = |9.5 − 8.6| =

(choose one) 0.5 / 0.9 / 1.6

The average of all moving ranges is denoted MR.

3. (LCL, UCL) for individual measurement chart

From the SAS graph, x̄¯ = (choose one) 5.2 / 10.5 / 10.6
The upper control limit and the lower control limit for the individual measure-
ment chart are,

x̄¯ ± 3σ̂ = x̄¯ ± 3R̄/d2

= x̄¯ ± 3MR/d2
= 5.2 ± 3(3.22069/1.128)
= (LCL, UCL) =

(choose one)
92 Chapter 5. Control Charts For Variables (ATTENDANCE 4)

(a) (−5.4, 11.7)

(b) (−3.4, 13.7)
(c) (−1.4, 15.7)

4. (LCL, UCL) for moving range chart

From the SAS graph, R̄ = (choose one) 3.22 / 5.52 / 10.6
The upper control limit and the lower control limit for the moving range chart
are, (choose one)

(a) (0, 10.52)

(b) (0, 11.72)
(c) (0, 15.75)

5. Difficulty with moving range charts

True / False
Since the moving range chart is calculated using two consecutive observations
in the individual measurements chart, the two charts are highly dependent on
one another. For example, a sharp increase between two values in the individual
chart will cause a sharp increase in the range chart. This dependence between
charts worries some analysts enough to not use the moving range chart.

5.5 Summary of Procedures For x̄, R and S Charts

This is covered in the previous three sections.

5.6 Applications of Variables Control Charts

There a number of interesting examples given in this section.