Chap 2 Introduction to Statistics

This chapter gives overview of statistics

including histogram construction,
measures of central tendency, and
dispersion
INTRODUCTION TO STATISTICS
 Statistics – deriving relevant information
from data
 Deals with
 Collection of data – census, GDP, football,
accident, no. of employees (male, female ,
department, etc)
 Collection , tabulation, analysis,
interpretation, an presentation of quantitative
data – can make some conclusions on sample
or population studied, make decisions on
quality
INTRODUCTION TO STATISTICS
 Use of statistics in quality deals with
second meaning. – inductive statistics
 Examples :
 What can we learn from the data?
 What conclusions can be drawn?
 What does the data tell about our process
and product performance? etc.
INTRODUCTION TO STATISTICS
 Understand the use of statistics vital
in business
 to make decisions based on facts
 in conducting business improvements
 in controlling and monitoring process,
products or service performance
 Application of statistics to real life
problems such as for quality
problems will result in improved
organizational performance
Collection of data
 Collect Data – direct observation or indirect
through written or verbal questions (market
research, opinion polls)
 Direct observation measured, visual checking,
classified as variables and attributes
 Variables data – measurable quality
characteristics
 Attributes – characteristics not measured but
classified as conforming or non-conforming
Collection of data
 Data collected with purpose
 Find out process conditions
 For improvement
 Variables – quality characteristics that are
measurable and countable
 CONTINUOUS - Dimensions, weight, height,
etc. (meter, gallon, p.s.i., etc.)
 DISCRETE - numbers that exhibit gaps,
countable, (no. of defective parts, no. of
defects/car, Whole numbers, 1, 2, 3….100)
Collection of data
 Attributes - quality characteristics that are non-
measurable and ‘those we do not want to
measure’
 Example : surface appearance, color,
Acceptable, non-acceptable conforming, non-
conf.
 Data collected in form of discrete values
 Variables (weight of sugar) CAN be classified as
attributes 
 weight within limits – number of conforming
 outside limits – no. of non conforming
 Summarizing Data
 Consider this data set on number of Daily Billing errors

0 1 3 0 1 0 1 0
1 5 4 1 2 1 2 0
1 0 2 0 0 2 0 1
2 1 1 1 2 1 1
0 4 1 3 1 1 1
1 3 4 0 0 0 0
1 3 0 1 2 2 3
 Data in this from
Meaningless
Not effective
Difficult to use
 Need to summarize data in the form of:
 Graphical – Freq. Dist., Histogram, Graphs,
Charts, Diagrams
 Analytical – Measures of central tendency,
Measure of dispersion
Frequency Distribution (FD)
 Summary of how data (observations) occur
within each subdivision or groups of observed
values
 Help visualize distribution of data
 Can see how total frequency is distributed
 Two types :
 Ungrouped data – listing of observed values
 Grouped data – lump together observed values
20.5  21.5 
 
 21.5  22.5 
22.5  23.5 
FD - Ungrouped Data
1. Establish array, No of Tally mark Frequency
arrange in ascending errors
0 /////////// 13
or descend (as in 1 ////
column 1) 2 /////
2. Tabulate the 3 ////

frequency – place 4
5
tally marking in
column 2
3. Present in graphical
form – Histogram,
Relative freq. distr.
FD – Ungrouped data
14
4 graphical representations
12
1. Frequency histogram 10
Frequency
8
2. Relative freq histogram 6
4
3. Cumulative frequency histogram 2
4. Relative cum frequency histogram 0 1 2 3 4 5

No error Freq Relative Cumulative Rel cum

freq freq freq
0 15 0.29 15 0.29
1 20 0.38 35 0.67
2 8 0.15 43 0.83
3 5 0.10 48 0.92
4 3 0.06 51 0.98
5 1 0.02 52 1.00
Total 52
Frequency Distribution For
Grouped Data
 Data which are continuous variable need grouping
Steps
1. Collect data and construct tally sheet
 Make tally - coded if necessary
 Too many data – group into cells
 Simplify presentation of distribution
 Too many cells – distort true picture
 Too few cells – too concentrated
 No of cells – judgment by analyst – trial and error
 Generally 5-20 cells
 Less than 100 data – use 5 –9 cells
 100 – 500 data – use 8 to 17 cells
 More than 500 – use 15 to 20 cells
 Cell interval (i)

CELL

Midpoint
UPPER BOUNDARY CELL NOMENCLATURE
2. Determine the range
 R = XH - XL
 R = range
 XH = highest value of data
 XL = lowest value of data
 Example :
 If highest number is 2.575 and lowest number
is 2.531, then
 R = XH - XL
 = 2.575 – 2.531
 = 0.044
3. Determine the cell interval
 Cell interval = distance between adjacent cell midpoints.
If possible, use odd interval values e.g. 0.001, 0.07,
0.5 , 3; so that midpoint values will have same no.
decimal places as data values.
 Use Sturgis rule.
 i = R/(1+ 3.322 log n)
 Trial and error
 h = R/i ;h= number of cells or cllases
 Assume i = 0.003; h = 0.044/0.003 = 15 cells
 Assume i = 0.005; h = 0.044/0.005 = 9 cells
 Assume ii = 0.007; h = 0.044/0/.007 = 6 cells
 Cell interval 0.005 with 9 cells will give best presentation
of data. Use guidelines in step 1.
4. Determine cell midpoints
 MPL = XL + i/2 (do not round)
 = 2.531 + 0.005/2 = 2.533
 1st cell have 5 different values (also the other
cells)

2.5 2.53
33 8
2.531 2.532 2.533 2.534
2.535
5. Determine cell boundaries
 Limit values of cell
 lower
 upper
 To avoid ambiguity in putting data
 Boundary values have an extra decimal
place or sig. figure in accuracy that
observed values
 + 0.0005 to highest value in cell
 - 0.0005 to lowest value in cell
6. Tabulate cell frequency
 Post amount of numbers in each cell
 Frequency distribution table

Cell boundary Cell MP Freq.

2.531 – 2.535 2.533 6
2.536 – 2.540 2.538 8
2.541 – 2.545 2.543 12
2.546 – 2.550 2.553 13
2.551 – 2.555 2.553 20
2.556 – 2.560 2.563 19
2.561 – 2.565 2.563 13
2.566 – 2.570 2.568 11
2.571 – 2.575 2.573 8
110
 Freq dist gives better view of central value and
how data dispersed than the unorganized data
sheet
 Histogram – describes variation in process
 Used to
 solve problems
 determine process capability
 compare with specifications
 suggest shape of distribution
 indicate data discrepancies, e.g. gaps
Characteristics Of Frequency
Distribution
 Symmetry, Number of modes (one, two
or multiple), Peakedness of data
Bi-modal

Sym.

Skew
Right
Skew
flatter
Left ‘very peak’
platykurtic
leptokurtic
Characteristics of Frequency
Distribution
 F.D. can give sufficient info to provide basis for
decision making.
 Distributions are compared regarding:-

Location Spread Shape

Descriptive Statistics
 Analytical method allow comparison between
data
 2 main analytical methods for describing data
 Measures of central tendency
 Measures of dispersion
 Measures of central tendency of a distribution -
a numerical value that describes the central
position of data
 3 common measures
 mean
 median
 mode
Measure of Central Tendency
 Mean - most common measure used
 What is middle value? What is average
number of rejects, errors, dimension of
product?
 Mean for Ungrouped Data - unarranged
 x (x bar)
n

X x i
x1  x 2   x n
 i1

n n
Mean
Example
A QA engineer inspects 5 pieces of a tyre’s
thread depth (mm). What is the mean thread
depth?
x1 = 12.3 x2 = 12.5 X3 = 12.0.
x4 = 13.0 x5 = 12.8
Σx i 62.5
x    12.5 mm
5 5
Mean - Grouped Data
 When data already grouped in frequency
distribution
h

f i xi
x  i 1
 Σfi 
fi (n)= sum. of freq.
fi = freq in the ith cell
n = no. of cells/class
xi = mid point in ith cell
Mean - Grouped Data
Cell (i) Class Mid Freq Fixi fi fixi
boundary Point (fi)
(xi)
1 1 – 20 10 2 20 2
2 21 – 40 30 10 300 12
3 41 - 60 50 20 1000 32
4 61 – 80 70 12 840 44
5 81 -100 90 6 540 50
Totals 2700

 fi xi
x  = 2700/50 = 54
fi
Weighted average
Tensile tests aluminium alloy conducted with different
number of samples each time. Results are as follows:
1st test : x1 = 207 MPa n=5 n
 w i xi
2nd test : x2 = 203 MPa n=6 xw  i 1
n
3rd test : x3 = 206 MPa n=3  wi
i 1

xw 
(5)(207)  6 (203)  3 (206)
 205 MPa
xw = weighted avg.
563 wi = weight of ith
or use sum of weights equals 1.00 average
W1 = 5/(5+6+3) = 0.36
W2 = 6/(5+6+3) = 0.43
W3 = 3/(5+6+3) = 0.21 Total = 1.00
Median – Ungrouped Data
 Median – value of data which divides total
observation into 2 equal parts
 Ungrouped data – 2 possibilities
 When total number of data (N) is a) odd or b)
even
 If N is odd ; (N+1/2)th value is median
 eg. 3 4 5 6 8 N+1/2=6/2=3 ,
3rd no.
 If N is even
 eg. 3 5 7 9 ½ of (5+7)=6

 NOTE: ORDER THE NUMBERS FIRST!

Median – Grouped Data
 Need to find cell / class having middle value &
interpolating in the cell using
n 
  cfm 
x 0.5  L m   2 i
 fm 
 
 
Lm = lower boundary of cell with the median
Cfm = Cum. freq. of all cells below Lm
fm =class/cell freq. where median occurs
i = cell interval
Example
MD = 40.5 + 10
= 53.5
Measures of dispersion
 describes how the data are spread out or
scattered on each side of central value
 both measures of central tendency & dispersion
needed to describe data
 Exams Results
 Class 1 – avg. : 60.0 marks
 highest : 95
 lowest : 25
 Class 2 – avg. : 60.0 marks
 highest : 100
 lowest : 15 marks
Measures of dispersion
 Main types – range, standard deviation,
and variance
 Range – difference bet. highest & lowest
value
 R = XH - XL
 Standard deviation
 Variance – standard deviation squared
 Large value shows greater variability or
spread
Standard deviation
 For Ungrouped Data
 s = sample std. dev.
n

  x  x
2
i
s i1
n 1
xi = observed value
x = average
n = no. of observed value
or use
2
 n
 n 
n x i    x i 
2

s i 1  i1 
n  n 1
 Standard deviation – grouped
data
Cell (i) Class Mid Freq Fixi fi fixi
boundary Point (fi)
(xi)
1 1 – 20 10 2 20 2
2 21 – 40 30 10 300 12
3 41 - 60 50 20 1000 32
4 61 – 80 70 12 840 44
5 81 -100 90 6 540 50
Totals 2700

   h 2
h
n fi x i     fi x i   50 (166,600)  (2700)2
2
  424.49  20.6
s 1  1  50  49 
n (n 1)

NOTE: DO NOT ROUND OFF fixi & fixi2

ACCURACY AFFECTED
Concept Of Population and Sample
 Total daily prod. of steel shaft. Population
 Year’s Prod. Volume of calculators
 Compute x and s sample statistics
 True Population Parameters
  and 
 Why sample? Sample
 not possible measure population
 costs involved
 100% manual inspection –
accuracy/error
Concept Of Population and Sample

SAMPLE POPN.
Statistics, Parameter
x , s  - mean
 - std. dev.
Normal Distribution
 Also called Gaussian distribution
 Symmetrical, unimodal, bell-shaped dist
with mean, median, mode same value
 Popn. curve – as sample size  cell
interval  - get smooth polygon

ND
Normal Distribution
 Much of variation in nature & industry
follow N.D.
 Variation in height of humans, weight of
elephants, casting weights, size piston
ring
 Electrical properties, material – tensile
strength, etc.
Example - ND
Characteristics of ND
 Can have different mean but same
standard deviation
 Different standard deviation but same
mean
Relationship between std
deviation and area under curve
Normal Distribution Example
 Need estimates of mean and standard
deviation and the Normal Table
 Example :
 From past experience a manufacturer
concludes that the burnout time of a
particular light bulb follows a normal
distribution. Sample has been tested and
the average (x ) found to be 60 days with
a standard deviation () of 20 days. How
many bulbs can be expected to be still
working after 100 days.
 Solution
 Problem is actually to find area under the curve beyond 100 days
 Sketch Normal distribution and shade the area needed
 Calculate z value corresponding to x value using formula
 Z=(xi - )/ = (100-60)/20 = +2.00
 Look in the Normal Table for z = +2.00 – gives area under curve as
0.9773
 But, we want x >100 or z > 2.00. Therefore Area = 1.000 – 0.9773
= 0.0227, i.e. 2.27% probability that life of light bulb is > 100
hours
σ =20

0 μ = 60 100 x
Test For Normality
 To determine whether data is normal
 Probability Plot - plot data on normal probability
paper
 Steps
1. Order the data
2. Rank the observations
3. Calculate the plotting position
i= rank , n=sample size,
100(i  position
PP= plotting 0. 5 ) in %
PP 
n
4. Label data scale
5. Plot the points on normal probability paper
6. Attempt to fit by eye ‘best line’
7. Determine normality
Example