Documente Academic
Documente Profesional
Documente Cultură
In practice, usually and are not known. In such case can be estimated by the
grand average that is average of sample means. Whereas, can be estimated using
distribution of relative range, W. The parameters of the distribution of W are a
function of the sample size of n.
The mean of W is d
2
. An estimator of is , which is related to R with following
relation.
2
R
d
=
Where, R is the average range of the m preliminary samples.
Using above relations, the control limits of X chart can be written as
X
2
X
2
2
2
2
3
UCL X R
d n
3
LCL X R
d n
andA isdefinedas
3
A
d n
= +
=
=
To define D
3
and D
4
, we need an estimator
R
in the R chart. Since, we have
assumed that the variable to be measured is normally distributed; we can estimate
R
using an eliminator
R
from the distribution of relative range.
We can write,
R W =
The standard derivation of W, represented by d
3,
is a known function of n.
So, the standard derivation of R is
R 3
d =
We can estimate
R
(because is unknown) with following relation
R 3
2
R
d
d
=
We can rewrite the control limits for R-chart by substituting the estimators of are
given below
R R
3
2
R R
3
2
3
3
2
3
3
2
UCL R 3
R
R 3d
d
LCL R 3
R
R 3d
d
where
d
D 1 3 and
d
d
D 1 3
d
= +
= +
=
=
=
= +
Since the factors A
2
, D
3
and D
4
are dependent on sample size, these can be drawn
from table 5.6.
TABLE 5.6: FACTORS FOR CALCULATING 3 LIMITS FOR THE X -
CHART AND R-CHART
Sample size
(n)
X - chart
factor (A
2
)
R
LCL factor
(D
3
)
R
UCL factor
(D
4
)
2 1.880 0 3.267
3 1.023 0 2.575
4 0.729 0 2.282
5 0.577 0 2.115
6 0.483 0 2.004
7 0.419 0.076 1.924
8 0.373 0.136 1.864
9 0.337 0.184 1.816
10 0.308 0.223 1.777
Steps to be followed to develop X chart and R chart
Step 1: collect data on the variable quality measurement and assign number to each
sample from 1 to m. (Preferable, m should be of atleast 20 samples)
Step 2: compute the range for each sample, where sample size for each sample is n.
compute the average range, R , for the set of samples, m.
R
1
=x
max
,
1
x
min
,
1
R
2
=x
max
,
2
x
min
,
2
.
.
.
.
R
1
=x
max
,
m
x
min
,
m
1 2 m
R R ........ R
R
m
+ + +
=
Where
m
R is the range for m
th
sample and x
max,m
and x
min,m
are the maximum and
minimum values of m
th
sample where the sample size is n.
Step 3: use the table for D
3
and D
4
and determine the control limits of the R-chart.
Step 4: Plot the sample ranges on the R-chart if all are in control then proceed to
next step. If all values are not in control then identify assignable causes. Correct
the assignable causes and repeat step 1, 2 and 3.
Step 5: Calculate X for each sample and determine X using following relation
1 2,1 n,1
1
1,2 2,2 n,2
2
1,m 2,m n,m
m
1 2 m
m
x x .............x
X
n
x x .............x
X
n
.
.
.
x x .............x
X
m
X X .............X
X
m
+
=
+
=
+
=
+
=
Where x
n,m
are the n
th
data in m
th
sample, where each sample is of sample size n.
Step 6: Use table to determine A
2
and control limits.
Step 7: Plot the sample means. If all the data points are in control, the process is in
statistical control in terms of process average and process variability. If any data
point is out of control then find assignable causes and return to step 1.
Sometimes, we can find the process exhibit sudden changes or patterns
which are not desirable ever though the data points lie within control limits. Such
instances can exhibit different patterns as shown in figure 5.18 and they need
investigation.
Figure 5.18 (a) presents normal behavior
Figure 5.18 (b) presents the data point out of control limits. Investigate for
assignable cause
Figure 5.18 (c) presents increasing or decreasing trend.
Figure 5.18 (d) presents a pattern called run, which is a sequence of
observations with a certain characteristic. If a run results due to five or more
observations then investigate their cause because in such case the probability is
low that such a result has taken place by chance.
Figure 5.18 (e) erratic behavior which needs to be monitored.
FIGURE 5.18: DIFFERENT PATTERNS OF PROCESS IN CONTROL BUT
NEEDS INVESTIGATION
Example 1:
A call-centre wants to check whether their service process is in control for the
quality measure of response time. The management expects to respond back to the
customer as soon as possible. Management decided to take hourly sample of four
calls to calculate a sample mean response time. The past records of call centre
working yield an estimated population mean response time of 4.5seconds with an
estimates range of 4 seconds. The data for sample means and sample range are
given below. Calculate the control limits for X chart and R chart and analyze the
pattern.
Time
(hrs)
Mean response
time (seconds)
Average Range of
response time
(seconds)
1 3 2
2 5 4
3 4 5
4 2 6
5 7 4
6 4 2
7 6 6
8 5 3
Solution:
The control limits for X chart with 8 samples of each sample of each sample size
of 4 can be calculated as given below.
2 X
2 X
2
X
X
UCL X A R
LCL X A R
A 0.729(fromtableforn 4)
R 4
UCL 4.5 (0.729 4) 7.416
LCL 4.5 (0.729 4) 1.584
= +
=
= =
=
= + =
= =
Similarly, we can calculate the control limits for R-chart
R 4
R 3
4
3
R
UCL D R
LCL D R
R 4
D 2.282(fromtableforn 4)
D 0(fromtableforn 4)
UCL 2.82 4 9.128
=
=
=
= =
= =
= =
We will plot X -chart
We can see that the process is in control as there is no data point out of control
limits.
The process is in control and shows normal behavior.
5.7.6 Attribute control charts or p-charts
p-charts are used for performance characteristic which is counted rather than
measured to control the proportion of defective services or product, p-charts are
used to
Accommodate unequal sample sizes.
Sample sizes are usually 50 or greater
Need 20-30 samples to construct the p-chart
Examples
Proportion of invoices with errors at any retailer.
Proportion of items requiring rework.
Proportion of incorrect saving account statements sent to customers.
Control limits for p-charts
A random sample is selected and that sample is inspected for each item relative to
the attribute in terms of yes or no decision. Calculate the sample proportion
defective, p.
p is the number of defective units divided by the sample size. Here, yes or no
means the output is defective or not. The underlying statistical distribution for such
attribute based random variables is based on binomial distribution. But, for large
sample sizes for such attribute measure, normal distribution is good approximation.
So, we need to define mean and standard deviation to this distribution.
p: average population proportion defective either determined from historical data
or a target value.
p
: the standard deviation of the distribution of proportion defectives, which can
be written as given below.
p
p(1 p)
n
=
Where n is the sample size.
The control limits for p-chart can be determined using following relations
p p
p p
UCL p Z
LCL p Z
= +
=
where Z is normal deviate.
Steps for plotting p-chart
Step 1: Take a random sample of sample size n.
Step 2: Count the number of defective services. Divide the number of defectives by
the sample size, n, to get a sample proportion defective p.
Step 3: Take m samples and determine p for all m samples and determine average
of p as p.
Step 4: Determine the control limits and plot sample proportion defectives on the
p-chart. The plots outside the control limits need investigation for assignable
causes.
Example:
A life insurance company registers many complaints from the customers regarding
the errors in customers profile at the time of policy issuing the company wants to
control the errors and for that samples are taken randomly for 10days with a
sample size of 30 applications. The average number of errors or proportion of error
for each day is given in the table below.
Day Number of
errors
1 5
2 3
3 7
4 10
5 2
6 3
7 6
8 4
9 3
10 4
Management wants to monitor the performance of policy using process using
control chart. Analyze the process.
Solution Using the sample data for 10 samples where each sample has sample size
of 30. We will calculate p
p
p p
p p
Total errors
p
Total numberof observations
5 3 7 ........ 4 47
0.157
(10)(30) 300
p(1 p)
n
0.157(1 0.157)
0.004
30
Thecontrol limitsare
UCL p Z
0.157 3(0.004)
0.17
LCL p Z
0.14
=
+ + + +
= = =
= =
= +
= +
=
=
=
Calculate the sample proportion errors as given in the following table and plot each
sample proportion error on the chart.
Day Sample proportion
error
1 5/30=0.167
2 3/30=0.1
3 7/30=0.233
4 10/30=0.333
5 2/30=0.067
6 3/30=0.1
7 6/30=0.2
8 4/30=0.133
9 3/30=0.1
10 4/30=0.133
The value lower than
p
LCL but greater than zero will be good for the company. The
system is out of control since four values are above
p
UCL .