You are on page 1of 16

EXPERIMENT 6: ATTRIBUTE CONTROL CHARTS

Agdigos, A.P., Galingana, C.L.T., Gualberto, S.R.S., Lopez, J.P. and Samaniego, K.P.
Abstract
The quality of Lemon Square Whatta Tops was evaluated by a quality control group called
Sayonara! Inc. The following are the defects that were identified: underweight, presence of cracked
chocolate layer, and deformity of cupcake. Sampling was done every 10 minutes until the 21st sampling
time; both for variable and constant sample size (n=20). Four control charts were constructed for the
number of nonconforming units (p and np), and number of nonconformities (u and c). Results show that u
and c charts both exhibit unstable processes. A pareto diagram of the defects was also constructed,
showing that the majority of the defects are either underweight or has a cracked chocolate layer. Both p
and np also indicated out of control processes. Lastly, no correlation was observed between the sample
size and the dispersion of both non-conforming units and non-conformities.
I.

Introduction
Control charts are used to monitor the quality of a process (Hung n.d.). There are two general
types of control charts: variable and attribute. In variable charts, characteristics are represented
numerically (NIST 2013). However, this type of data collection may be impractical or unsuitable. When
that is the case, attribute control charts are used (NIST 2013). In attribute control charts, a characteristic
whether passes or fails, or is present or absent (Minitab 2016). Attribute control charts plot the
defects/nonconformities or the defectives/nonconforming units. A nonconformity or a defect refers to
characteristics, whereas a nonconforming unit or a defective refers to a unit of the product itself (Minitab
2016).
There are two general types of attribute control charts: one is concerned with the proportion of
nonconforming items relative to the number of items inspected, and the other is concerned with the
number of nonconformities per item (Hung n.d.). In the first category, the p- and np-charts are used. In the
second, it is the c- and u-charts (McNeese 2009). The p-chart (p stands for proportion) plots the fraction
of the defectives in a sample of n items, with the control limits based on a binomial distribution. The
sample size may vary each time (McNeese 2009). The np-chart plots the number of defectives in a sample
of n items per time period, with the control limits based on a binomial distribution. Unlike in p-charts, the
sample size has to be the same each time (McNeese 2009).
The c-chart (c stands for count) plots the number of defects over time. Like the np-chart, the
sampling area must be the same each time. Lastly, the u-chart plots the number of defects per unit over
time. Similar to the p-chart, the sampling area may vary each time (McNeese 2009). Both the c- and ucharts are based on a Poisson distribution (Minitab 2016).
In the food industry, attribute control charts can be used to evaluate the performance of a specific
process or step at a shift or time interval (Udupa 1978). If there is an unusual trend in the number of
defects or defectives, it may indicate that the specific process needs maintenance (Udupa 1978). Attribute
control charts are also easier to interpret, allowing even personnel to understand them (Besterfield 2009).
This exercise aims to demonstrate how attribute control charts are constructed and interpreted. It
also aims to demonstrate the effect of sample size to the dispersion of defects and defective units.

II.

III.

Methodology
The manufacturer of Lemon Square Whatta Tops wanted to have their chocolate cupcakes
evaluated by a quality control group called Sayonara! Inc. The group decided to evaluate the quality of
the product according to weight, presence of a cracked chocolate layer, and deformity of cupcake. For
every 10 minutes, cupcake samples (variable sample size) are obtained and classified into either
conforming or nonconforming.
From the defectives (nonconforming), the presences of defects were identified: underweight,
cracked chocolate layer, deformed cupcake. Defects may be of several types, and the deformity of
cupcake was counted as 1 to 3, depending on the severity. Underweight cupcakes were those that did not
conform to the declared net weight. Evaluation of the presence of cracked chocolate layer was done by
visual inspection. Lastly, deformity of cupcakes were counted as 1=trace, 2=moderate, and 3=severe.
The data were recorded and plotted to create u and p charts. The total number of defects per
subgroup (u) and total defectives per subgroup (p) for each subgroup were calculated. The subgroup u and
p were plotted onto the u and p chart, respectively. Then, the central line (CL), upper control limit (UCL),
and lower control limit (LCL) for each chart were determined by Sayonara! Inc. These control lines were
drawn onto each chart. Finally, the control charts were analyzed by the quality control group to determine
if the process is in control.
The same procedure was done in the control charts for constant sample size, except that exactly
20 cupcake samples were obtained every 10 minutes. A c chart was constructed instead of a u chart, and
an np chart instead of a p chart.
Summary of Results
Control charts for the data in tables 1B to 4B of the appendix are shown in figures 1 to 4
respectively. It can be observed in figure 1 that two points (sample number 5 and 14) lie beyond the
control limits (see appendix for calculated control limits) deeming the process out of control. No other
trends were observed.

Figure 1. Control chart for fraction nonconforming with variable sample size

In figure 2, there is no point lying beyond the control limits (CL=10.4286; UCL=17.1306; LCL=
3.727) however points 8 to 11 showed a constant decrease in number of nonconforming units and point 15
to 18 are shown to be hugging the central line hence the process is out of control.

Figure 2. Control chart for number of nonconforming units with constant sample size (n=20)
For the non-conformities, the u-chart in figure 3 showed three points (sample number 1, 10 and
13) are lying beyond the UCL while sample numbers 16 to 19 are close to the center line indicating that
the process is out of control. This conclusion can also be drawn from the c-chart in figure 4 due to sample
number 1 lying above the UCL.

Figure 3. Control chart for no. of non-conformities with variable sample size

Figure 4. Control chart for number of nonconformities with constant sample size (n=20)
Pareto diagram using the defects data shows that ~45% of defects are underweight, and that
~40% of defectives are damaged chocolate layer.

Figure 5. A pareto diagram concerning the defects of Lemon square Whatta tops
Plotting the sigma values from table 1B and 3B of the appendix against the sample size results to
a plot with line equation y=-0.0008x + 0.1086 and y = -0.0015x + 0.2021 respectively, both having
correlation coefficient of 0.9565 hence a negative correlation. The plots are shown below.

(a)

(b)
Figure 6. A plot showing the (a) effect of sample size on variation of non-conforming and (b)effect of
sample size on the variation of non-conformities
IV.

Interpretation of Results
For this study, a total of four control charts were constructed in order to define the quality of
Lemon Square Whatta Tops. Two control charts were designed, each, for the number of nonconforming
units (p and np), and number of nonconformities (u and c).
Based on the p chart, where some points lie beyond the control limits, the process can be
considered as unstable, due to some special causes present in the process. However, it does not entirely
mean that the process is faulty. Points below the LCL may indicate that the process performance has
improved or is continuously improving (Walker et al. 2009). Results from the np-chart show hugging the
center line which may indicate that the variability within the process has decreased. Decreased
variability is important in SQCs, so it is recommended to find out what has happened and recalculate the
control limits to reflect the reduced variability (Kelley 1999).
Control charts for the number of nonconformities, u and c charts, both exhibit unstable processes.
Since there are only 1-2 points that lie beyond the upper control limit, it can be assumed that the plotted
point(s) is in error, so it is less likely that the process has deteriorated (Walker et al. 2009).
6

The quality of the product is also important, regardless of the amount of defectives, in order to
stay competitive in the business. Nonconformities charts (U and C) can be used to check the quality of a
specific product. Combined with a pareto diagram, nonconformities charts can be used to decide which
specific part of the product they can change to improve the product the most using the least amount of
resources (Griffith 1996). A pareto diagram of the defects, as shown in Figure 5, shows that the majority
of the defects are either underweight or damaged chocolate layer. Problems causing these defects should
focused upon as they get more reward for effort used (Griffith 1996). Pareto diagrams can also be used
with defectives rather than defects to check which products needs more improvement than the others.
Variable sample size is sometimes unavoidable especially in small companies where they cannot
afford to have a larger and consistent sample size. U-charts and P-charts can be used to address variable
sample size for defects and defectives, respectively. U-charts utilizes the normalized number of defects, in
the sense that it uses the number of defectives per sub-group as shown in the following equation
(Ishikawa 1976).

sample
x total number of nonconformities /defects
u= =

n
P-chart also works in a similar manner, as it uses the proportion of defectives per sub-group
(Ishikawa 1976).

sample
Di number of defectives
Pi= =

By dividing the number of defects or defectives by the sample size for each subgroup, u-charts
and p-charts transform an integer data into a percentage data, effectively normalizing the effects of
varying sample size.
As stated by Besterfield (2009), np charts are easier to understand for working personnels than
the p chart. This can be easily argued by the fact that the np chart uses a fixed number of sample,
compared to the p chart. The control limits are also fixed, which allows the personnel to see what is the
objective of their work. It also gives the exact number of defectives in a given sub-group, allowing the
personnel to easily estimate how well they are doing.
Several factors affect the spread of the data points in attribute control charts, one of which is the
sample size. Since control limit formulas depend on sampling variation, according to Laney (2002) larger
sample sizes makes the limit squeeze in towards the center line, causing overdispersion. Increasing
sample sizes improves precision and somehow shows the invalidity of the binomial assumption (Laney,
2002). This is evident in figures 6a and 6b, where the sample variance has a negative correlation with the
sample size.
In summary, all the control charts (p, np, u and c) indicated out of control process however it does
not automatically mean that the process is faulty and some points may be due to errors during the
inspection. The pareto chart showed that the company should address the problem that may be causing the
production of underweight cupcakes. Lastly, the inspection showed the negative correlation between the
sample size and the dispersion of both non-conforming units and non-conformities.

V.

References
7

Besterfield D. 2009. Quality control: chapter 8 - control charts for attributes. Upper Saddle River. Pearson
Education. 540p.
Griffith G. 1996. Statistical process control methods for long and short runs. ASQ Quality Press. 250p.
Hung C. n.d. Control Charts for Attributes: Some Variations [Internet]. Iowa State University; [cited 2016
Apr 12]. Available from http://www.public.iastate.edu/~wrstephe/HungCC.pdf
Ishikawa K. 1967. Guide to quality control. Michigan. Asian Productivity Organization. 226 p.
Kelley D. 1999. Measurement made accessible. Thousand Oaks, Calif.: Sage Publications. 88p.
Laney D. 2002. Improved control charts for attributes. Quality Engineering 14:531-537.
McNeese B. 2009. Attribute Control Charts Overview [Internet]. SPC for Excel; [cited 2016 Apr 12].
Available
from
https://www.spcforexcel.com/knowledge/attribute-control-charts/attributecontrol-charts-overview
Minitab. 2016. Attribute control charts in Minitab [Internet]. Minitab.com; [cited 2016 Apr 12]. Available
from
http://support.minitab.com/en-us/minitab/17/topic-library/quality-tools/
controlcharts/understanding-attributes-control-charts/attributes-control-charts-in-minitab/
NIST. 2013. e-Handbook of Statistical Methods [Internet]. NIST/SEMATECH; [cited 2016 Apr 12].
Available from http://www.itl.nist.gov/div898/handbook/pmc/section3/pmc33.htm
Udupa KS. 1978. Statistical Quality Control Charts on Attributes in Fish Processing Technology. Fish.
Technol. 15(1978): 109-13.
Walker H, Elshennawy A, Gupta B, McShane-Vaughn M. 2009. The certified quality inspector handbook.
Milwaukee, Wis.: ASQ Quality Press. 442p.

APPENDIX A
RAW DATA
Table 1A. Number of nonconforming and nonconformities in chocolate cupcakes (Lemon square Whatta
tops) evaluated with variable sample size

Student
Agdigos
Belen
Benig
Buligan
De
Guzman
Galingan
a
Gualbert
o
Leister
Lopez
Melo
Minguill
o
Padilla
Pestio
Regis
Sabocoj
an
Samanie
go
Sanchez
Sapuay
Sayson
Tobias
Yan

Dama
ged
chocol
Underwe
ate
ight
layer
29
12
19
15
16
6
8
13

Defor
med
cupca
ke
14
0
7
2

# Nonconforming
units, Di
31
20
25
21

n
42

47
39
37

12

35

13

21

36

55

21
13
14
25

14
14
15
15

14
0
12
13

41
22
31
31

56

10
18
37
9

22
10
29
6

6
1
3
0

29
28

51

40
14

51
39

10

13

29

13
7
7
46
17
13

23
21
22
3
7
17

6
3
12
7
4
0

39
28
29

51
43
45

48
24

69
45

30

46

33
40
43

40

Table 2. Number of nonconforming and nonconformities in chocolate cupcakes (Lemon square Whatta
tops) evaluated with constant sample size (n=20)

Student

Underw
eight

Dama
ged
choco
late
layer

Defor
med
cupca
ke

#
Nonconforming
units, Di
9

Agdigos
Belen
Benig
Buligan
De
Guzman
Galingan
a
Gualbert
o
Leister
Lopez
Minguillo
Melo
Padilla
Pestio
Regis
Sabocoja
n
Samanie
go
Sanchez
Sapuay
Sayson
Tobias
Yan

9
5
3
3

6
4
6
6

12
0
2
1

12
13
10
10

12

6
8
7
1
7
3
10
3

4
8
4
6
4
10
3
5

5
0
5
2
5
1
0
0

12
15
11
9
7
13
13
8

10

5
9
5
6
7
3

6
2
6
2
3
5

3
0
3
1
1
0

11
11
11
8
10
8

APPENDIX B
PROCESSED DATA
Table 1B. Data for p-chart

Student

# Non-conforming
units, Di

ni

Agdigos

31

42

Belen

20

47

Benig

25

39

Buligan

21

37

pp i
(Di/ni)
0.7380
95
0.4255
32
0.6410
26
0.5675
68

Standa
rd
deviati
on
i
0.0748
12
0.0707
21
0.0776
36
0.0797
07

UCL
0.8466
29
0.8343
55
0.8551
01
0.8613
13

LCL
0.3977
55
0.4100
28
0.3892
82
0.3830
7
10

De
Guzman

12

35

Galingana

36

55

Gualberto

41

56

Leister

22

33

Lopez

31

40

Melo

31

43

Minguillo

29

51

Padilla

28

40

Pestio

40

51

Regis

14

39

Sabocojan

13

29

Samanieg
o

39

51

Sanchez

28

43

Sapuay

29

45

Sayson

48

69

Tobias

24

45

Yan

30

46

TOTAL

592

936

0.3428
57
0.6545
45
0.7321
43
0.6666
67
0.775
0.7209
3
0.5686
27
0.7
0.7843
14
0.3589
74
0.4482
76
0.7647
06
0.6511
63
0.6444
44
0.6956
52
0.5333
33
0.6521
74
13.066
03

0.0819
53
0.0653
76
0.0647
89
0.0844
0.0766
6
0.0739
37
0.0678
91
0.0766
6
0.0678
91
0.0776
36
0.0900
32
0.0678
91
0.0739
37
0.0722
76
0.0583
68
0.0722
76
0.0714
86

0.8680
5
0.8183
19
0.8165
6
0.8753
91
0.8521
71
0.8440
04
0.8258
65
0.8521
71
0.8258
65
0.8551
01
0.8922
89
0.8258
65
0.8440
04
0.8390
18
0.7972
95
0.8390
18
0.8366
49

0.3763
33
0.4260
65
0.4278
24
0.3689
93
0.3922
12
0.4003
8
0.4185
18
0.3922
12
0.4185
18
0.3892
82
0.3520
95
0.4185
18
0.4003
8
0.4053
65
0.4470
88
0.4053
65
0.4077
35

p-bar = 0.6222 (Center line)


Table 2B. Data for np-chart (n=20)

Student

# non-conforming units,
Di

pi

Agdigos

12

0.6

Belen

13

0.65

Benig

10

0.5

Buligan

10

0.5

UCL
LCL
17.130 3.7265
61
31
17.130 3.7265
61
31
17.130 3.7265
61
31
17.130 3.7265
61
31
11

De
Guzman
Galingan
a
Gualbert
o

0.25

12

0.6

12

0.6

Leister

15

0.75

Lopez

11

0.55

Melo

0.45

Minguillo

0.35

Padilla

13

0.65

Pestio

13

0.65

Regis
Sabocoja
n
Samanie
go

0.4

10

0.5

11

0.55

Sanchez

11

0.55

Sapuay

11

0.55

Sayson

0.4

Tobias

10

0.5

Yan
TOTAL

0.4
10.95

17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61
17.130
61

3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31
3.7265
31

p-bar = 0.521429
n(p-bar) = 10.42857 (Center line)
Table 3B. Data for u-chart

Student

Sampl
e
size,
n

Total
number of
nonconformiti
es, x

u (# of
nonconformit
ies per unit)

Agdigos

42

55

1.30952

Belen

47

34

0.7234

Standa
rd
deviati
on
i
0.0601
6
0.0568
7

UCL

LCL

1.2143
79
1.1917
32

0.3861
61
0.4088
07
12

Benig

39

29

0.74359

Buligan
De
Guzman

37

23

0.62162

35

15

0.42857

Galingana

55

40

0.72727

Gualberto

56

49

0.875

Leister

33

27

0.81818

Lopez

40

41

1.025

Melo

43

53

1.23256

Minguillo

51

38

0.7451

Padilla

40

29

0.725

Pestio

51

69

1.35294

Regis

39

15

0.38462

Sabocojan
Samanieg
o

29

16

0.55172

51

42

0.82353

Sanchez

43

31

0.72093

Sapuay

45

41

0.91111

Sayson

69

56

0.81159

Tobias

45

28

0.62222

Yan
TOTAL

46

30

936

731

0.65217
16.8057

0.0624
31
0.0640
97
0.0659
02
0.0525
72
0.0521
0.0678
7
0.0616
46
0.0594
57
0.0545
95
0.0616
46
0.0545
95
0.0624
31
0.0724
0.0545
95
0.0594
57
0.0581
2
0.0469
37
0.0581
2
0.0574
85

1.2300
11
1.2414
73
1.2539
03
1.1621
44
1.1588
99
1.2674
48
1.2246
05
1.2095
35
1.1760
68
1.2246
05
1.1760
68
1.2300
11
1.2986
27
1.1760
68
1.2095
35
1.2003
37
1.1233
53
1.2003
37
1.1959
65

0.3705
28
0.3590
67
0.3466
36
0.4383
95
0.4416
41
0.3330
92
0.3759
34
0.3910
04
0.4244
72
0.3759
34
0.4244
72
0.3705
28
0.3019
13
0.4244
72
0.3910
04
0.4002
02
0.4771
86
0.4002
02
0.4045
75

= 0.813034188 (Center line)


Table 4B. Data for c-chart (n=20)

Student

Total
c-bar
number
of nonconformit

UCL

LCL

13

ies,
c
Agdigos

27

Belen

Benig

11

Buligan
De
Guzman
Galinga
na
Gualber
to

10

Leister

16

Lopez
Minguill
o

16

Melo

16

Padilla

14

Pestio

13

5
15
15

Regis
Sabocoj
an
Samani
ego

12

Sanchez

11

Sapuay

14

Sayson

Tobias

11

Yan

14

12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38
12.52
38

23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05
23.14
05

1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111
1.907
111

14

APPENDIX C
SAMPLE CALCULATIONS
A. p-chart

^p i( Agdigos)=

pbar=

Di
ni

total ^p i
number of samples

31
=0.738095
42

13.06603
21

= 0.6222

( pbar ) (1 pbar)
n
( 0.6222 ) (10.6222)
UCL = 0.6222+3
= 0.846629
42
( 0.6222 )(10.6222)
LCL = 0.62223
= 0.397755
42
Control limits =

pbar 3

B. np-chart
same calculation of pbar as in A
Center line = n(pbar) = 20 x 0.521429 = 10.42857
Control limits = Center line 3 ( pbar ) (1 pbar)
UCL = 10.42857 +3 ( 0.521429 ) (10.521429) = 17.1306
LCL = 10.42857 3 ( 0.521429 ) (10.521429) = 3.72653
C. u-chart

sample
x total number of nonconformities
u= =

n
55
u ( Agdigos )= =1.3095
42
761
=
= 0.813034
936
Center line =
Control limits =
UCL =
LCL =

0.813034
=1.230432813
n 42
0.813034
0.8130343
= 0.395636
n 42
0.813034+3

D. c-chart
15

Center line = c-bar = average of total number of non-conformities


Control limits cbar 3 cbar
UCL = 12.524+3 12.524=23.141
LCL = 12.5243 12.524= 1.90711

16

You might also like