1/29/2021

CONCEPT OF VARIATION CONCEPT OF VARIATION

&
DATA COLLECTION
• No two items are ever made exactly alike.

• This variation is universal and it occurs in

Ushani Pallearachchi every manufacturing/service process.
Deputy Director (Training)

KINDS OF VARIATION NATURAL OR INHERENT VARIATION

 This variation occurs due to random causes.

In a manufacturing process, the Examples:
variation of quality characteristics is an
 Slight variation of raw materials (though within
inevitable phenomenon, and it consists the specification)
of two types.
 Slight vibration of machines
 Lack of human perfection in reading
1. Natural or Inherent variation. instruments and setting controls.
2. Unnatural or External variation.  The variation due to the particular random
cause is very small. But, many random causes
act together to yield a substantial total.

UNNATURAL OR EXTERNAL VARIATION VARIATION contd…

 This variation occurs due to any assignable or identifiable cause.
Examples:
 Some assignable causes arrive gradually.
 Tool wear
 Any one assignable cause can result in a large
 Machine wear amount of variation.
 Some assignable causes arrive suddenly.
 Tool damage
 Batch of defective raw materials  The presence of assignable causes can be
 Some assignable causes are due to regular change detected and eliminated.

 Variation in temperature
 Variation in humidity

1
 1/29/2021

STATISTICALLY CONTROLLED PROCESS FACTORS AFFECTING VARIATION

1. Man
The process operating with only
2. Material
inherent (Natural) variation is said
3. Method
to be in the state of Statistically
4. Machines – 4M
Control.

QUALITY CHARACTERISTIC

• Characteristic which shows variation is called

variable.
Data Collection
• A variable which can be used to indicate the
quality of a product or process is called
quality characteristic.

DATA Data collection

• Data is the numerical (Quantitative)/non • Source
numerical (Qualitative) expression of quality – Process
characteristic. – Lot or batch
Continuous or • 100% inspection
Measurable data – Collection of data from each and every
Numerical
Eg. Weight, Length, Tensile strength
data item of a process or lot
Discrete or Countable • Sampling
data – Collection of data using one or more items selected
Eg. Number of defects, from a lot or process
Number of defectives

2
 1/29/2021

Importance of Sampling Data collection contd…

• Advantages of sampling
100% inspection Sampling – Cost
– Time
Economy Expensive & Economical – Accuracy
uneconomical – Easy to handle
Accuracy Poor due to Can be – Only feasible in case of destructive testing
tiredness satisfactory for a
given population
Destructive Not feasible Only feasible
testing (Possible) alternative

PURPOSE OF DATA COLLECTION PURPOSE OF DATA COLLECTION

To make decision or take action about

the population, on the basis of data
 For various analysis
collected from it.
 Defects Analysis
 Action on a process  Correlation study
 Process Control
 Process Analysis
 Action on a Lot
 Lot Inspection
 Estimation of Lot Quality

POPULATION, SAMPLE AND DATA POINTS TO BE CONSIDERED WHEN

COLLECTING DATA
Define objective clearly.
Sample Data
Population Find suitable procedure for data collection.
Action on Process Lot Sample Data
process Apply stratification in collecting data.
Process Control Action
Process Analysis

Action on Take reliable measurement.

Lot Sample Data
lot
Inspection Action Find right way to record data.
Estimation of lot
quality

3
 1/29/2021

Preparation of Frequency Distribution

Table
Data can be summarized in
several different forms; • Step 1: Count number of data and find
Minimum and Maximum observations.

1. Tabular methods • Step 2: Find the Range. Range = Max. – Min.

2. Graphical methods
3. Quantitative methods • Step 3: Determine number of Classes for the
distribution.

i) Decide the number of classes. ii) Calculate the Class Interval

Number of Recommended number of
observations classes Range
20 - 50 6 Class Interval 
Number of Classes
51 - 100 7

101 - 200 8 Round this result to some convenient

number (preferably the nearest uneven
201 - 500 9 number with the same number of significant
digits as the actual data).
501 - 1000 10

Over 1000 11 - 20

iii) Construct the classes by listing class 02 Graphical Summarization

boundaries.
Histogram
– The class boundaries should have one A histogram is constructed by drawing vertical
more significant digit than the actual data bars whose areas are proportional to the
corresponding frequencies.
and should be end in 5.
The width of bars is proportional to the class
– The class interval should be constant interval.
throughout the entire frequency distribution.
The class interval is constant throughout the
iv) Tally each observation into the appropriate frequency distribution.
class and then list the total frequency (f) for each The height of each bar is proportional to the
class. corresponding frequency.

4
 1/29/2021

The histogram for the data vs. moisture Ogive

content of a product Frequency distribution can be described
diagrammatically by a frequency polygon.
HISTOGRAM OF MOISTURE CONTENT
A cumulative frequency distribution can be
18
represented by a curve called an Ogive.
16

14

12
Frequency

OGIVE OF MOISTURE CONTENT

10
60
8

Cumulative Frequency
50
6
40
4

2 30

0 20
8.815-9.245

9.245-9.675

9.675-10.105

10.105-10.535

10.535-10.965

10.965-11.395

11.395-11.825
10

0
9.245 9.675 10.105 10.535 10.965 11.395 11.825

Moisture Content
Moisture Content

3 Quantitative Method (a) Measures of Central Tendency

There are several ways to measure the

In quantitative methods, data are central tendency of data.
summarized in two ways.
1.Mean
a) Measures of Central Tendency 2.Median and
b) Measures of Dispersion 3.Mode

Calculation of Arithmetic Mean for

Mean or Arithmetic Mean (Average)
Frequency Distribution

If x1, x2, x3……….xn are n observations or data,

f1x1  f 2 x2  f 3 x3  .......... f n xn 1 n
then mean x   f i xi
n n i 1
Sum of n observatio ns
x
n
where, xi is the Class Mid value and fi is the frequency of
class.
x1  x2  x3  .......... xn 1 n
x   xi
n n i 1

5
 1/29/2021

Example : Calculate the mean of the data given Population Mean ( x ) or ( )

Class Mid value (xi) fi fixi

9.030 3 27.090
9.460 9 85.140

9.890 16 158.240 x1  x2  ............  x N

10.320 11 113.520 x
10.750 9 96.750 N
11.180 1 11.180
11.610 1 11.610
where, N is the number of observations in the population.
Total 50 503.530

x
f x i i

503.530
 10.070
f i 50

Median Calculation of Median from Frequency

Distribution
When the observations are arranged in order of magnitude, the
median is defined as the middle value, if the number of
n F 
Md  L  2 c
observation is odd. If the number of observation is even, the
median is
defined as the mean of two middle values.
 f 
 
120 228 229 231 231 232 233 234 237 237
where L = the lower class boundary of the median class
Two middle values n = the total number of observations
F = the cumulative frequency of all classes lower than
Mean of two middle values is the Median = 231.5.
the median class
231 231 232 233 237 f = the frequency of the median class
c = the class interval
Median = 232

Comparison of Mean and Median

But, we can estimate the Mode using the following formula:

Mean Median where:
•L is the lower class boundary of the modal group
1. Based on all Based on one or two •fm-1 is the frequency of the group before the modal group
•fm is the frequency of the modal group
observations observations •fm+1 is the frequency of the group after the modal group
•w is the group width
2. Easy to compute Something easy to
compute
3. Can be handled Hard to handle
mathematically mathematically
fm − fm-1
4. Has relatively Sampling error is Estimated Mode = L + ×w
(fm − fm-1) + (fm − fm+1)
small sampling larger than mean if
error population is normal
5. Affected by Not seriously
extreme values affected by extreme

6
 1/29/2021

(b) Measures of Dispersion

The mean and the median of all three sets are
identical and equal to 300. But, the amount of
In order to indicate a set of data using summary variation of each set of data is different
measure, at least two measures are necessary.
There are three commonly used measures of
Set A = 100, 200, 300, 400, 500 dispersion.
Set B = 50, 150, 300, 450, 550
Set C = 250, 300, 300, 300, 350 1. Range
2. Variance
3. Standard deviation

01 Range 02 Variance
Variance is a measure of deviation about the mean of a set of
The range is simply the difference between the largest and
the smallest value in the set of data. Use generally when the data.
number of data is small. Population variance is defined by

Example : For the three sets given above, the ranges are as
follows.
N
R = 500 – 100 = 400
 x  x
2
R = 550 – 50 = 500 i
R = 350 – 250 = 100   2 i 1
N

2
03 Standard Deviation n
 n 
 xi   xi 
2

i 1  i 1  n
The positive square root of variance is called the standard s
deviation. Population standard deviation is defined as, n 1
Calculation of sample standard deviation from a frequency distribution
N

 xi  x 
2
2
n
 n 
  fx   f i xi 
i 1 2

N i 1
i i
 i 1  n
s
n 1
Sample standard deviation is defined as,
n

 x  x
2 The standard deviation of a distribution indicates the
i variability of the data (quality characteristics). If the

s  i 1 standard deviation is very small, then the data are

more reliable and consistent.
n 1