You are on page 1of 13

Communications in Statistics—Theory and Methods

ISSN: 0361-0926 (Print) 1532-415X (Online) Journal homepage: https://www.tandfonline.com/loi/lsta20

A Nonparametric Synthetic Control Chart Using


Sign Statistic

S. K. Khilare & D. T. Shirke

To cite this article: S. K. Khilare & D. T. Shirke (2010) A Nonparametric Synthetic Control Chart
Using Sign Statistic, Communications in Statistics—Theory and Methods, 39:18, 3282-3293, DOI:
10.1080/03610920903249576

To link to this article: https://doi.org/10.1080/03610920903249576

Published online: 02 Sep 2010.

Submit your article to this journal

Article views: 305

View related articles

Citing articles: 31 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=lsta20
Communications in Statistics—Theory and Methods, 39: 3282–3293, 2010
Copyright © Taylor & Francis Group, LLC
ISSN: 0361-0926 print/1532-415X online
DOI: 10.1080/03610920903249576

A Nonparametric Synthetic Control Chart


Using Sign Statistic

S. K. KHILARE1 AND D. T. SHIRKE2


1
Department of Statistics, Vivekanand College, Kolhapur, India
2
Department of Statistics, Shivaji University, Kolhapur, India

This article presents a synthetic control chart for detection of shifts in the process
median. The synthetic chart is a combination of sign chart and conforming run-
length chart. The performance evaluation of the proposed chart indicates that the
synthetic chart has a higher power of detecting shifts in process median than the
Shewhart charts based on sign statistic as well as the classical Shewhart X-bar
chart for various symmetric distributions. The improvement is significant for shifts
of moderate to large shifts in the median. The robustness studies of the proposed
synthetic control chart against outliers indicate that the proposed synthetic control
chart is robust against contamination by outliers.

Keywords Average run length; CRL chart; Nonparametric control chart; Sign
statistic.

Mathematics Subject Classification Primary 62G86; Secondary 62P30.

1. Introduction
Control charts are used to monitor shifts in the center of the distribution. Most of
the control charts are based on the sample means when observations of samples are
taken sequentially under the normality condition. If the distribution of observations
is non normal then the central limit theorem is usually used to justify the assumption
that the distribution of sample mean is approximately normal. The nonparametric
control charts are used for monitoring the process median (or mean) are based
on the signs computed within samples and used in place of sample means in the
Shewhart chart. The chart is labeled to be the nonparametric chart if in-control
average run length (ARL) does not depend on the underlying process distribution.
In case of charts based on signs, the ARL will be same for all distributions for which
median equals to the target value. In nonparametric control charts the assumption
of normality is not necessary for calculating the control limits. Another advantage
is that the nonparametric control charts are usually more efficient than the charts
based on X-bar when the distribution of the observations is heavy tailed, that is

Received April 18, 2009; Accepted August 7, 2009


Address correspondence to S. K. Khilare, Department of Statistics, Vivekanand College,
Tarabai Park, Kolhapur 416003, India; E-mail: shashi_khilare@rediffmail.com

3282
Nonparametric Synthetic Control Chart 3283

when observations in the tails of the distribution have a higher probability than for
normal distribution (see, for example, Amin et al., 1995). In nonparametric control
charts variance of the process need not be known or estimated in order to apply the
control chart. In fact, the nonparametric control charts for controlling median are
not affected by changes in the variance as long as location parameter  is constant.
The nonparametric control charts may be particularly useful when a process is just
starting up and it is desirable to apply a control charts before there is a enough data
to get a reasonable estimate of variance and/or assess the normality of the process.
In quality control applications McGilchrist and Woodyer (1975) proposed a
distribution-free cumulative sum technique for use in monitoring rainfall amounts.
Lehmann (1975) studied the nonparametric statistical methods based on ranks.
Bakir and Reynolds (1979) have proposed nonparametric cumulative sum chart
using the Wilcoxon signed-rank statistic. Park and Reynolds (1987) developed
nonparametric Shewhart and CUSUM charts based on linear placement statistics.
Amin and Searcy (1991) studied the behavior of the EWMA control chart using the
Wilcoxon signed-rank statistic. Amin et al. (1995) have studied the nonparametric
quality control charts based on the sign statistic. Chakraborti et al. (2001),
Bakir (2006), Chakraborti and Eryilmaz (2007) proposed control charts based on
signed-rank statistic. Chakraborti and Van de Wiel (2008) have proposed Mann–
Whiteny statistic based control chart. Das and Bhattacharya (2008) developed a
nonparametric control chart for controlling variability.
This article presents a synthetic control chart based on sign statistic for
monitoring the process median and compare the performance of synthetic chart with
Shewhart charts using X-bar and sign statistic in terms of out-of-control ARL.

2. Shewhart Charts Using Sign Test Statistic


Suppose the characteristic of interest X follows distribution with cumulative
distribution function (cdf) F. We assume that F is a cdf of a continuous
random variable with median . Let 0 be a target value of the median. A sample
of n observations is taken at a regular time interval from the process. Let Xi =
Xi1  Xi2      Xin  be the sample taken at the ith time point. At any time t, each
observation from the sample is compared with target value 0 and the number of
observations above 0 and below 0 is recorded.
Define


1 if Xij > 0
signXij − 0  = 0 if Xij = 0 (1)


−1 if Xij < 0 

where Xij is the jth observation in the ith sample.


Since the distribution of observations is assumed to be continuous, PXij − 0 =
0 = 0. In practice occasional zero may occur which can be signed alternatively +1
and −1.
Let
n

SNi = signXij − 0  i = 1 2 3     (2)
j=1
3284 Khilare and Shirke

where SNi is the difference between number of observations above 0 and number
of observations below 0 in the ith sample. A random variable Ti = SN2i +n gives
the number of positive signs in the sample of size n and has binomial distribution
with parameters n and p, where p = pXij > 0 . As long as median remains at 0 ,
we have p = p0 = 1/2. That is, pXij > 0  = pXij < 0  = 1/2 and ESNi  = 0. The
chart signals that shift has occurred if SNi  ≥ a, where a > 0 is a specified constant.
The chart signals that shift has occurred in the positive direction if SNi ≥ a and
chart signals that the shift has occurred in the negative direction if SNi < a. Using
SNi has the advantage of keeping control limits symmetric about zero, while Ti has
advantage of being a standard binomial random variable.
The sign test is very simple to apply, since the numerical values of observations
need not be determined exactly, only we have to see whether the observation is
above 0 or below 0 . Thus, this test procedure is very useful when little is known
about the distribution of observations or when the distribution is known to be non-
normal. Although since magnitude of the observations are not used, the SNi will be
relatively inefficient as compared to the sample mean of ith time point, when the
observations are nearly normal.
For sign chart, ARL denote the average number of samples required to
signal a process shifts and ARL for two-sided control chart is defined as

1
ARL =  (3)
P
where P is a detecting power of sign chart.

P = P SNi  ≥ a
P = 1 − P SNi  < a
P = 1 − P −a < SNi < a
 
−a + n a+n
P = 1−P < Ti <
2 2
   
a+n −a + n
P = 1 − P Ti < + P Ti <
2 2

The ARL for one-sided control limit in the positive direction is

1
ARL =  (4)
PSNi ≥ a

The largest possible in-control ARL values of symmetric one-sided and two-sided
control chart are 2n and 2n−1 , respectively, when p = 1/2 and SNi = n. Unless n is
of a moderate size, it may be difficult to achieve even approximately a specified in-
control ARL(0). In fact, the largest possible value of ARL(0) when n = 10 is 1024,
which is obtained when a = 10.
The average run length for non normal process observations of X-bar chart is
given as

1
ARL =  (5)

P X ≥ 0 + a 
Nonparametric Synthetic Control Chart 3285

where “a” is fixed and the distribution of X-bar is known. If the interest is in one
direction only, then one of the control limits 0 + a or 0 − a is used.
In this article, we propose a control chart which combines sign chart due to
Amin and conforming run length chart due to Bourke. In the following, we first
discuss conforming run length (CRL) chart.

3. The Conforming Run Length Control Chart


The conforming run length chart is proposed by Bourke (1991). The conforming run
length is the number of inspected units between two consecutive non conforming
units including ending non conforming unit. In Fig. 1, the white and black circles
denote the conforming and non conforming units, respectively. Suppose process
start at t = 0, then the three samples of CRL are displayed. CRL1 = 4, CRL2 = 5,
CRL3 = 3. The idea behind the CRL chart is that the conforming run length will
change when the fraction non conforming in a process p changes. Namely, the CRL
is shortened as p increases and is lengthened as p decreases.
The random variable CRL follow a geometric distribution. The probability
mass function of CRL is

PCRL = p1 − pCRL  CRL = 1 2 3     (6)

The cumulative probability function and mean value of CRL are, respectively,

FCRL = 1 − 1 − pCRL
(7)
1
CRL = 
p

If CRL is less than lower control limit L of CRL chart , then an upward process
shift is signaled. Therefore, for detection of an upward process shift (increase in p),
a single lower control limit L of CRL chart is sufficient and L can be derived from
Eq. (7), we have:

CRL = FL = 1 − 1 − pL

Figure 1. Conforming run length.  – Conforming unit; • – Nonconforming unit.


3286 Khilare and Shirke

(8)
ln1 − CRL 
L= 
ln1 − p0 

where CRL is the Type-I error probability of the CRL chart and p0 is the in-control
fraction non conforming. L must be rounded to an integer. If a sample CRL is a
less than or equal to the L, then the fraction non conforming p has increased and
out-of-control status will be signaled.
For the CRL chart, ARLCRL , is the average number of CRL samples required
to detect out-of-control fraction non conforming p is given by

1
ARLCRL =
CRL
(9)
1
ARLCRL = 
1 − 1 − pL

Finally, let ANICRL be the average number of the inspected units required to signal
a fraction non conforming shift and be equal to the product of CRL and ARLCRL .

ANICRL = CRL × ARLCRL


(10)
1 1
ANICRL = × 
p 1 − 1 − pL

For CRL chart, if a CRL value falls between lower and upper control limits of the
CRL chart, then the process is considered to be under control. However, if CRL
value is less than the lower control limit of CRL chart, then upward process shift
is signaled and if CRL value greater than upper control limit of CRL chart, then
downward process shift is signaled. The presentation of CRL chart usually based
on the 100% inspection, because every unit has to be accounted for and classified as
either conforming unit or non conforming one.

4. Synthetic Control Chart Using Sign Statistic


In the literature, Wu and Spedding (2000) studied the synthetic control chart for
detecting small shifts in the process mean and Wu et al. (2001) studied the synthetic
control chart for fraction non conforming and reported that the synthetic control
chart has higher power of detecting out-of-control signal. Wu and Spedding (2001)
developed the synthetic control charts for attributes. Davis and Woodall (2002)
discussed the runs rule representation of the synthetic control charts to identify
shifts in the process mean using X-bar based procedure. In the following, we discuss
the synthetic control chart using sign statistic.
A synthetic control chart is proposed by combining the operations of the
sign chart and conforming run length chart. The operations of synthetic chart are
outlined below.
(1) Determine sign chart based “a” (>0), sample size n and the CRL based lower
control limit L.
(2) Take a sample of n units for inspection and calculate SNi .
Nonparametric Synthetic Control Chart 3287

(3) If SNi < a, a sample is a conforming one and control flow goes back to step (2).
Otherwise, a sample is a non conforming one and control flow continues to the
next step.
(4) Check number of samples between the current and previous non conforming
samples. This number is taken as CRL value for synthetic chart.
(5) If CRL value satisfy CRL ≥ L, then the process is said to be under control and
control flow goes back to the step (2). Otherwise, the process is taken as out-of-
control and control flow continues to the next step.
(6) Take action to locate and remove the assignable causes. Then go back to
step (2).
Generally, the synthetic chart works the same way as the ordinary CRL chart,
except that each unit in the CRL chart is replaced by a sample of n units in the
synthetic chart. In a CRL chart, the fraction non conforming p is the probability
that non conforming unit occurs. However, in synthetic chart P corresponds to the
non conforming sample occurs.
It should be noted that, a SNi value larger than “a” in the synthetic chart does
not signal an out-of-control status immediately. Instead of giving out-of-control
signal, the sample is marked as non conforming sample only. The out-of-control
status is signaled only when the CRL value below the CRL based lower control
limit L.
In the synthetic chart, ARLs denote the average number of SNi samples
required to detect a shift in the median. In the attribute chart, the fraction non
conforming p is the probability that a non conforming unit occurs. In the synthetic
chart, this probability should be replaced the probability that a non conforming SNi
sample occurs; which is the probability that a sample falls beyond the control limits
of the sign chart. Thus, the out-of-control average run length for synthetic control
chart can be calculated from Eq. (10) by replacing fraction non conforming p by
detecting power P of sign chart

1 1
ARLS  = ×  (11)
P 1 − 1 − PL

where, P is the detecting power of the sign chart calculated as follows:

P = PSNi ≥ a
 
a+n
P = P Ti ≥ 
2

Since the equal positive and negative shifts in the median will result in the same
amount of increase of P, only the positive shifts in the median will be discussed.
Furthermore, since P never decreases, the CRL chart only needs a lower control
limit.
The synthetic chart is designed to minimize ARLs. If  = 0, then in-control
ARL is
1 1
ARLS 0 = × (12)
P0 1 − 1 − P0L

The above equation can be used to determine two design parameters L and “a” for
fixed sample size n in synthetic chart. If ARLs(0) is set at the value specified by the
3288 Khilare and Shirke

Table 1
Different sets of values of L and “a” for
n = 10 and specified in-control ARL(0) = 1024
L a ARLs(0)
8 8 1124.627
8 9 1124.627
9 8 1004.998
9 9 1004.998
10 8 909.312
10 9 909.312

user, then any set of values of L and “a” satisfying Eq. (12) will result in a synthetic
chart that can meet the requirement for ARLs(0) or false alarm rate.
In Tables 1 and 2, specified in-control ARL(0) depends only on sample size n.
For the proposed synthetic control chart specified in-control ARL(0) = 1024 for
the case n = 10 and ARL(0) = 32 for n = 5. In-control ARL(0) is usually small
if n is small. In Tables 1 and 2, values of ARLS (0) are depend on n and specified
constant “a”. ARLs(0) remains the same for certain values of n even when “a” is
changed. It is seen from the following example. Let a = 8 then a+n 2
= 8+10
2
= 9 and
if a = 9 then 2 = 2 = 95. ARLS 0 is unchanged for both the choices of “a”
a+n 9+10

as the underlying random variable takes only integer values.


The problem is to find a pair L a such that ARLs(0 ) is equal to the specified
value and ARLs() is minimum. Thus, we can define an optimization problem
in which the objective function is to minimize the ARLs() subject to equality
constraint function for ARLs(0). L is the only independent design variable and the
variable “a” is a function of L. The general idea of the synthetic chart is to identify
a pair of L and “a”, so that ARLs is minimized, while ARLs(0) is approximately
equal to specified value.
The complete design procedure for the synthetic chart can be outlined as below.
(1) Specify sample size n, and ARL(0).
(2) Initialize L as 1 and 1 ≤ a ≤ n.
(3) Calculate ARLs from the current “a” and L, using Eq. (11).
(4) If ARLs was improved (reduced), increase L by one and go back to step (3).
Otherwise, go to the next step.

Table 2
Different sets of values of L and “a” for
n = 5 and specified in-control ARL(0) = 32
L a ARLs(0)
1 3 28.444
1 4 28.444
2 3 15.693
2 4 15.693
Nonparametric Synthetic Control Chart 3289

(5) If ARLs is approximately equal to the specified in-control ARL(0), take the
current L and “a” as the final values in synthetic chart.
The proposed synthetic control chart is very simple to apply since the numerical
values of observations need not be determined exactly, only we have to see whether
the observation is above median or below median. This test procedure is very simple
when little is known about distribution of observations or when distribution is
non normal. Thus the chart is as easy as sign chart. Such a control chart would be
easy from a users perspective.

5. Performance of Nonparametric Synthetic Control Chart


Using Sign Statistic
Tables 3–5 give the ARL values for Shewhart charts using SNi , X-bar, and synthetic
chart for normal, double exponential, and Cauchy distributions for the case n = 10.
For normal and double exponential distributions, scale parameter is adjusted to give
= 1. The density of X-bar for double exponential distribution is given in Johnson
and Kotz (1970) and cumulative distribution can be obtained from the density by
integration. In normal distribution, out-of-control average run length of synthetic
chart is half or less as compared to the SNi for the shifts between 0.10–1.50 and its
performance is best over the X-bar. In double exponential and Cauchy distributions,
out-of-control average run length of synthetic chart is half or less as compared to
the X-bar and SNi for the shifts between 0–1.25. The improvement of synthetic chart
is significant in all cases. For Cauchy distribution, distribution of X-bar is same as
the distribution of Xi , therefore, Shewhart X-bar chart is almost insensitive to mean
shifts under this distribution.

Table 3
ARL values for one-sided Shewhart X-bar, sign and
synthetic charts for normal distribution when n = 10
and ARL(0)= 1024 and = 1
Shifts X-bar SNi Synthetic
 − 0  a = 09794 a = 10 a = 9 L = 9
0.00 102400 102400 100499
0.10 36895 47582 30027
0.20 14584 23511 10334
0.25 9488 16899 6398
0.40 2989 6836 1888
0.50 1544 4002 999
0.60 869 2469 603
0.75 427 1307 346
1.00 190 563 195
1.25 124 305 140
1.50 105 199 116
1.75 101 151 106
2 100 126 102
3290 Khilare and Shirke

Table 4
ARL values for one-sided Shewhart X-bar, sign and
synthetic charts for double exponential distribution
when n = 10 and ARL(0)= 1024 and = 1
Shifts X-bar SNi Synthetic
 − 0  a = 105 a = 10 a = 9 L = 9
0.00 102400 102400 100499
0.10 44664 27622 13125
0.20 20202 11915 3922
0.25 13790 7672 2186
0.40 4688 2671 652
0.50 2429 1694 428
0.60 1329 1056 295
0.75 605 669 217
1.00 230 365 153
1.25 134 244 126
1.50 108 185 113
1.75 102 153 106
2.00 100 135 103

6. Robustness of the Synthetic Control Chart against Outliers


In practice, there are situations in which process is stable (in-control), although
a few outliers occasionally contaminate its output. A single extreme outlying
observation may trigger an out-of-control signal while in fact the process is

Table 5
ARL values for one-sided Shewhart X-bar, sign and
synthetic charts for Cauchy distribution when n = 10
and ARL(0)= 1024 and = 1
Shifts X-bar SNi Synthetic
 − 0  a = 849 a = 10 a = 9 L = 9
0.00 102585 102400 100499
0.10 102465 12532 4205
0.20 102344 3128 766
0.25 102283 1931 479
0.40 102102 759 235
0.50 101981 523 187
0.60 101860 403 162
0.75 101679 307 140
1.00 101377 233 124
1.25 101075 196 116
1.50 100773 175 111
1.75 100471 162 108
2.00 100169 152 106
Nonparametric Synthetic Control Chart 3291

in-control; thus increasing the false alarm rate and decreasing the in-control ARL
of the control chart. In this section we are interested in studying the effect of
distributional contamination on the proposed synthetic chart. For this study we
consider normal distribution. A standard model for generating normally distributed
process that produce occasionally outliers is the contaminated normal distribution.
The contamination normal distribution is a mixture of two normal
distributions. The cdf of contaminated normal distribution is

Gx =
Fx   + 1 −
Fx  1 (13)

where Fx   is the cdf of normal distribution with parameters  and 2 and 0 ≤

≤ 1. We will refer
and 2 as the percentage of contamination and extremity of
contamination, respectively. When  = 0, the process is in-control though producing
occasional outliers. When
= 0 and  = 0, Eq. (13) becomes cdf of standard normal
distribution. The in-control ARL values for proposed non parametric synthetic
chart are computed using 10,000 simulations when subgroup sample size is n =
10. The simulations are made for all possible combinations of  2 
, where 2 =
4 9 16 and
= 00 001 005 010 015, and 020.
Tables 6 and 7 present simulated in-control values for Shewhart X-bar chart
and sign based synthetic chart for various levels of contamination when the
underlying process distribution is normal. The value of constant “a” is obtained to
give in-control ARL = 1024, when process is operating under normal distribution
with no outliers.
Following are the important findings from Table 6.
(1) Under very low percentage
= 1 % and light extremity of contamination 2 =
4, outliers have a large effect as the ARL of Shewhart X-bar chart drops to
290 which entails about 3.5 times as many false alarms as the expected ARL of
1024. For moderate extremity of contamination 2 = 9, outliers have noticeable
effect as the ARL of Shewhart X-bar chart drops to 64, which entails about 16
times as many false alarms as the expected ARL of 1024. When extremity of
contamination 2 = 16, outliers have substantial effect as the ARL of Shewhart
X-bar chart drops to 38, which entails about 27 times as many false alarms as
the expected ARL of 1024.

Table 6
In-control ARL values of Shewhart X-bar chart for
stable process with occasionally outliers for
contaminated normal distribution. n = 10 a = 09794
Extremity of outliers
Percentage of
contamination 2 = 4 2 = 9 2 = 16

= 00 1028 1018 2026

=1 290 64 38

=5 72 15 9

= 10 34 8 5

= 15 23 6 4

= 20 17 5 3
3292 Khilare and Shirke

Table 7
In-control ARL values of sign based synthetic control
chart for stable process with occasionally outliers for
contaminated normal distribution. n = 10 a = 9
Extremity of outliers
Percentage of
contamination 2 = 4 2 = 9 2 = 16

= 00 1028 1024 1015

=1 1019 1018 1024

=5 1026 1035 1016

= 10 1021 1029 1029

= 15 1007 1034 1038

= 20 1031 1026 1013

(2) Under the usual percentage


= 10% and extremity of contamination 2 =
4 9 16, outliers have a substantial effect as the ARL of Shewhart X-bar chart
drops to 34, 8, and 5, respectively, which entails about 30, 128, and 205 times as
many false alarms as the expected ARL of 1024.
(3) When the percentage of contamination is as large as
= 20% and extremity
of contamination 2 = 4 9 16, outliers have a substantial effect as the ARL of
Shewhart X-bar chart drops to 17, 5, and 3, respectively, which entails about 60,
128, and 340 times as many false alarms as the expected ARL of 1024.
In the simulation study of robustness, Table 6 shows that in-control ARL of
Shewhart X-bar chart changes significantly depending on the level of contamination.
This is because Shewhart X-bar chart is based on actual observations in the sample
and hence affected by the outliers. On the other hand, Table 7 show that in-control
ARL values of the proposed sign based synthetic control chart are unaffected by
contamination. This is because signs are not affected by outliers. This property of
the proposed chart based on sign statistic is very attractive.

7. Conclusion
In this article, we presented the operation, design, and performance of the
nonparametric synthetic control chart for monitoring process median and studied
the robustness of the control chart against contamination by outliers. The synthetic
chart is a combination of sign chart and conforming run length chart. The
ARL performance of the proposed synthetic control chart is compared with the
Shewhart X-bar chart and sign chart under normal, double exponential, and
Cauchy distributions. The ARL comparisons under normal, double exponential,
and Cauchy distributions indicate that the synthetic chart has a greater detecting
power for shifts in the process median than the Shewhart charts using SNi and
X-bar. The improvement in the ARL is more significant for small to moderate
shifts and for large shifts ARL declines. Our simulation studies indicate that
the proposed synthetic control chart is robust against contamination by outliers
while Shewhart X-bar chart is not. Therefore, to monitor process operating under
normal distribution, the nonparametric synthetic control chart using sign statistic
is recommended to use, because this chart is more efficient than Shewhart X-bar
Nonparametric Synthetic Control Chart 3293

chart and also robust against contamination by outliers. The improvement is more
significant in the all cases and out-of-control ARL is likely to be reduced by at least
half. The ARL values are calculated under the zero-state mode and steady state
behavior of the proposed chart is under investigation and will be reported elsewhere.

Acknowledgments
The authors would like to thank the two reviewers for their useful and detailed
comments which contributed to the improvement of this article.

References
Amin, R. W., Searcy, A. J. (1991). A nonparametric exponentially weighted moving average
control schemes. Commun. Statist. Simul. Computat. 20:1049–1072.
Amin, R. W., Reynolds, M. R. Jr., Bakir, S. T. (1995). Nonparametric quality control charts
based on the sign statistic. Commun. Statist. Theor. Meth. 24(6):1597–1623.
Bakir, S. T., Reynolds, M. R. Jr. (1979). A nonparametric procedure for process control
based on within-group ranking. Technometrics 2:175–183.
Bakir, S. T. (2006). Distribution-free quality control charts based on signed-rank-like
statistic. Commun. Statist. Theor. Meth. 35:743–757.
Bourke, P. D. (1991). Detecting shift in fraction nonconforming using run length control
charts with 100% inspection. J. Qual. Technol. 23:225–238.
Chakraborti, S., Eryilmaz, S. (2007). A nonparametric Shewhart-type signed-rank control
chart based on runs. Commun. Statist. 36:335–356.
Chakraborti, S., Van de Wiel, M. A. (2008). A nonparametric control charts based on
Mann–Whitney statistic. IMS Collection 1:156–172.
Chakraborti, S., Van de Laan, P., Bakir, S. T. (2001). Nonparametric control charts: An
overview and some results. J. Qual. Technol. 33:304–315.
Davis, R. B., Woodall, W. H. (2002). Evaluating and improving the synthetic control chart.
J. Qual. Technol. 34(2):200–208.
Das, N., Bhattacharya, A. (2008). A new nonparametric control chart for controlling
variability. Qual. Technol. Quant. Manage. 5(4):351–361.
Johnson, N. L., Kotz, S. (1970). Continuous Univariate Distributions-2. Boston: Houghton
Mifflin Company.
Lehmann, E. L. (1975). Nonparametrics: Statistical Methods Based on Ranks. San Francisco:
Olden-Day.
McGilchrist, C. A., Woodyer, K. D. (1975). Note on a distribution-free CUSUM technique.
Technometrics 17:321–325.
Park, C., Reynolds, Jr. M. R. (1987). Nonparametric procedure for monitoring a location
parameters based on linear placement statistic. Sequen. Anal. 6:303–323.
Wu, Z., Spedding, T. A. (2000). A synthetic control chart for detecting small shifts in the
process mean. J. Qual. Technol. 32:32–38.
Wu, Z., Spedding, T. A. (2001). Implementing synthetic control charts. J. Qual. Technol.
32:32–38.
Wu, Z., Yeo, S. H. (2001). Implementing synthetic control charts for attributes. J. Qual.
Technol. 33:113–115.
Wu, Z., Yeo, S. H., Spedding, T. A. (2001). A synthetic control chart for detecting fraction
nonconforming increases. J. Qual. Technol. 33:104–111.

You might also like