View
53
Download
0
Category
Tags:
Preview:
Citation preview
3/11/2012
1
Copyright Tapan P Bagchi 1
Sessions 3 and 4
Probability and Statistics in Six Sigma: A review
Copyright Tapan P Bagchi 2
Why Study PROBABILITY?
Occurrence of defects in production is stochastic—such events cannot be exactly predicted. In decisions about such events we rely on the theory of probability. When our decisions require data analysis, the typical methods are obtained from statistics.
3/11/2012
2
Copyright Tapan P Bagchi 3
Applications of Probability Theory and Statistics in Business
Forecasting, Inventory Management Quality Assurance Project Risk Management Investment Portfolio Design Business simulation, Market research Game Theory and Strategy Formulation Six Sigma
Copyright Tapan P Bagchi 4
The Scope of Sessions 3 and 4
• Experiments Outcomes, Events and Sample Spaces
• What is probability?
• Basic Rules of Probability
• Probabilities of Compound Events
• An introduction to Distributions
• Test of hypothesis and inference
3/11/2012
3
Copyright Tapan P Bagchi 5
Learning Objectives
Understand the concepts of sample space and probability distribution and
construct sample spaces and distributions in simple cases
conditional probability and independent events; understand how to compute the probability of a compound event
Use simulations to construct empirical probability distributions and to make informal inferences about the theoretical probability distribution
Copyright Tapan P Bagchi 6
What is Probability?
Probabilitythe study of chance associated with the occurrence of random or stochastic events
Types of Probability Classical (Theoretical) Relative Frequency (Experimental)
3/11/2012
4
Copyright Tapan P Bagchi 7
Classical Probability
Rolling dice and tossing a coin are activities associated with a classical approach to probability. In these cases, you can list (or enumerate) all the possible outcomes of an experiment and determine the actual probabilities of each outcome.
Copyright Tapan P Bagchi 8
Sample Space, Events and RVsThe possible outcomes of a stochastic or random process are called events.
An event is a deterministic process has only one possible outcome.
The probability of a particular event is the fraction of outcomes in which the event occurs. The probability of event A is denoted by P(A).
Random variables map events to numbers. AS
4.3 X
3/11/2012
5
Copyright Tapan P Bagchi 9
Events may be (a) mutuallyexclusive, or (b) independent
Probability of an event remains between 0 (the event never occurs) and 1 (the event always occurs).
Two events are mutually exclusive if occurrence of one precludes the occurrence of the other.
Events whose occurrence do not depend on the occurrence of any other events are called independent events.
Copyright Tapan P Bagchi 10
Experiments, Outcomes, Events and Sample Spaces
Experiment: An experiment is any activity from which resultsare obtained. A random experiment is one in which the
outcomes, or results, cannot be predicted with certainty.
Examples:1. Flip a coin2. Flip a coin 3 times3. Roll a die4. Draw a random sample of size 50 from a population
Trial: A physical action , the result of which cannot be predetermined
3/11/2012
6
Copyright Tapan P Bagchi 11
Basic Outcomes and Sample SpacesBasic Outcome (o): A possible outcome of the experiment
Sample Space: The set of all possible outcomes of an experiment
Example: A company has offices in six cities, San Diego, Los Angeles, San Francisco, Denver, Paris, and London. A new employee will be randomly assigned to work in on of these offices.
What are the Outcomes?
What is the Sample Space?
Copyright Tapan P Bagchi 12
Assigning Probabilities to Events
Probability of an event P(E): “Chance” that an event will occur
• Must lie between 0 and 1• “0” implies that the event will not occur• “1” implies that the event will occur
Types of Probability:
Objective Relative Frequency Approach Equally-likely Approach
Subjective – based on beliefs, judgment and past experience
3/11/2012
7
Copyright Tapan P Bagchi 13
OddsIf the odds that an event occurs is a:b, then
( ) aP Aa b
Example: If the odds of the horse “Chetak” winning the Hong Kong Derby are 9:2, what is the subjective probability that he will win?
Copyright Tapan P Bagchi 14
Probabilities of Events
Let A be the event A = {o1, o2, …, ok}, where o1, o2, …, ok are k different outcomes. Then
1 2( ) ( ) ( ) ( )kP A P o P o P o
Problem: The number on a license plate is any digit between 0 and 9. What is the probability that the first digit is a 3? What is the probability that the first digit is less than 4?
3/11/2012
8
Copyright Tapan P Bagchi 15
• Start with the Law of Complements: “If A is an event, then the complement of A, denoted by ,represents the event composed of all basic outcomes in S (the sample space) that do not belong to A.”
A = set of outcomes thatmake event A
• By Additive Law of Probability: P(A) + P( ) = 1
Probabilities of Compound Events
A
A
A
S = set of all outcomes
A
Copyright Tapan P Bagchi 16
“If A is an event, then the complement of A, denoted by ,represents the event composed of all basic outcomes in S that do not belong to A.”
Law of ComplementsA
A
SLaw of Complements:
Example: If the probability of getting a “working” computer is 0.7,What is the probability of getting a defective computer?
( ) 1 ( )P A P A
3/11/2012
9
Copyright Tapan P Bagchi 17
• Unions of Two Events“If A and B are events, then the union of A and B, denoted by AB, represents the event composed of all basic outcomes in A or B.”
• Intersections of Two Events“If A and B are events, then the intersection of A and B, denoted by A∩B, represents the event composed of all basic outcomes in A and B.”
Unions and Intersections of Two Events
S
BA
Copyright Tapan P Bagchi 18
Additive Law of ProbabilityLet A and B be two events in a sample space S. The probability of the union of A and B is
( ) ( ) ( ) ( ).P A B P A P B P A B
S
BA
3/11/2012
10
Copyright Tapan P Bagchi 19
Using Additive Law of Probability
S
CM
Example: At Cornell, all first-year students must take chemistry and math. Suppose 15% fail chemistry, 12% fail math, and 5% fail both. Suppose a first-year student is selected at random. What is the probability that student selected failed at least one of the courses?
Copyright Tapan P Bagchi 20
Mutually Exclusive EventsMutually Exclusive Events: Events that have no basic outcomes in common, or equivalently, their intersection is the empty set .
S
BA
Let A and B be two events in a sample space S. The probability of the union of two mutually exclusive events A and B is
( ) ( ) ( ).P A B P A P B
3/11/2012
11
Copyright Tapan P Bagchi 21
Multiplication Rule and Independent EventsTwo independent events can occur together!
Multiplication Rule for Independent Events: Let A and B be two independent events, then
( ) ( ) ( ).P A B P A P B
Examples:• Flip a coin twice. What is the probability of observing two heads?
• Flip a coin twice. What is the probability of getting a head and then a tail? A tail and then a head? One head?
• Three computers are ordered. If the probability of getting a “working” computer is .7, what is the probability that all three are “working” ?
Copyright Tapan P Bagchi 22
Definitions
Sample Space - the list of all possible outcomes from a probabilistic experiment. 3-Children Example:
S = {BBB, BBG, BGB, BGG, GBB, GBG, GGB, GGG}
Each individual item in the list is called a Simple Event or Single Event.
3/11/2012
12
Copyright Tapan P Bagchi 23
Probability of Single Events with Equally Likely Outcomes
If each outcome in the sample space is equally likely, then the probability of any one outcome is 1 divided by the total number of outcomes.
outcomes ofnumber total1event) simple(
outcomes,likely equally For
P
Copyright Tapan P Bagchi 24
Three Children Example Continued
A couple wants 3 children. Assume the chance of a boy or girl is equally likely at each birth.
What is the probability that they will have exactly 3 girls?
What is the probability ofhaving exactly 3 boys?
3/11/2012
13
Copyright Tapan P Bagchi 25
Probability of Combinations of Single Events
But in general, an Event can be a combination of Single Events.
The probability of such an event is the sum of the individual probabilities.
Copyright Tapan P Bagchi 26
Three Children Example Continued
P(exactly 2 girls) = __P(exactly 2 boys) = __P(at least 2 boys) = __P(at most 2 boys) = __P(at least 1 girl) = __P(at most 1 girl) = __
Sample space =
3/11/2012
14
Copyright Tapan P Bagchi 27
Types of Probability
Classical (Theoretical)
Relative Frequency (Experimental, Empirical)
Copyright Tapan P Bagchi 28
Relative Frequency Probability
Uses actual experience to determine the likelihood of an outcome.
What isthe chanceof scoringa B or better?
Grade Frequency
A 20
B 30
C 40
Below C 10
3/11/2012
15
Copyright Tapan P Bagchi 29
Empirical Probability
Given a frequency distribution, the probability of an event, E, being in a given group is
nxP
ondistributi in the sfrequencie totalgroup theoffrequency E)(
Copyright Tapan P Bagchi 30
A Problem:Two-way Tables and Probability
Find:P(M)
P(A)
P(A and M)
Given: Got AGot < A
Total
Male 30 45
Female 60 65
Total
To solve this problem we need some theory.
3/11/2012
16
Copyright Tapan P Bagchi 31
Probability Fundamentals What is wrong with the statements? The probability of rain today is -10%. The probability of rain today is 120%. The probability of rain or no rain today is 90%.
1) (1)(0)(
spacesamplePeventPeventP REMEMBER!
These are the axioms of Kolmogorov.
Copyright Tapan P Bagchi 32
Probability Rules
Let A and B be events
Complement Rule:P(A) + P( not A) = 1
3/11/2012
17
Copyright Tapan P Bagchi 33
Set Theory Notation
Union: A or B (inclusive “or”)
BA
BA
Intersection: A and B
Copyright Tapan P Bagchi 34
Probability Rules
Union P(AUB) = P(A or B)
)()()()( BAPBPAPBAP
3/11/2012
18
Copyright Tapan P Bagchi 35
Two-way Tables and Probability
Find:P(M)P(A)P(A and M)P(A if M)
Got A Got < A
Total
Male 30 45 75
Female 60 65 125
Total 90 110 200
Copyright Tapan P Bagchi 36
Conditional Probability
P(A|B) = the conditional probability of event A happening given that event B has happened
“probability of A given B”
)()()|(
BPBAPBAP
3/11/2012
19
Copyright Tapan P Bagchi 37
Independent Mutually events exclusive
events
Head TailKGP Student had bicycle
stolen
It rained in Mumbai on
July 15
Car accidents
Obama gets elected
Fire in lab
Copyright Tapan P Bagchi 38
Independence
Events A and B are “Independent” if and only if
)()|( APBAP
Using the data in the two-way table, can you say that getting an “A” grade is independent of the student’s being male?
3/11/2012
20
Copyright Tapan P Bagchi 39
Two-way Tables and Probability
Question:Are grades and
gender independent?
Got A Got < A
Total
Male 30 45 75
Female 60 65 125
Total 90 110 200
Copyright Tapan P Bagchi 40
Terminology
The sum of probabilities of all mutually exclusive events in a process is 1. For example, if there are n possible mutually exclusive outcomes, then
P( i) 1i1
n
3/11/2012
21
Copyright Tapan P Bagchi 41
Simple probabilities
If A and B are mutually exclusive events, then the probability of either A or B to occur is the union
P(A B) P(A) P(B)
Copyright Tapan P Bagchi 42
Simple probabilitiesIf A and B are independent events, then the probability that both events A and B occur is the intersection
P(A B) P(A) P(B)Example: The probability that a US president is bearded is ~14%, the probability that a US president died in office is ~19%, thus the probability that a president both had a beard and died in office is ~3%. If the two events are independent, 1.3 bearded out of 43 presidents are expected to fulfill the two conditions. In reality, 2 died. (A close enough result.)
3/11/2012
22
Copyright Tapan P Bagchi 43
Conditional probabilities
What is the probability of event A to occur given than event B did occur. The conditional probability of A given B is
P(A | B) P(A B)
P(A)Example: The probability that a US president dies in office if he is bearded 0.03/0.14 = 22%. Thus, out of 6 bearded presidents, 22% (or 1.3) are expected to die. In reality, 2 died. (Again, a close enough result.)
Copyright Tapan P Bagchi 44
Joint Probability
For events A and B, joint probability Pr(AB) stands for the probability that both events happen.
Example: A={HH}, B={HT, TH}, what is the joint probability Pr(AB)?
Think—can any outcome lead to the occurrence of A and also B? Can A and B occur together?
3/11/2012
23
Copyright Tapan P Bagchi 45
Independence
Two events A and B are independent in casePr(AB) = Pr(A) Pr(B)
Independence does not mean that the events A and B cannot occur together
A set of events {Ai} is independent in case
Pr( ) Pr( )i iii A A
Copyright Tapan P Bagchi 46
The birth of a son or a daughter are mutually exclusive events.
Events—birth of a daughterdaughter and the birth of a child with AB+ blood typeAB+ blood type——are are not mutually exclusive (they are independentevents).
3/11/2012
24
Copyright Tapan P Bagchi 47
Independence
Two events A and B are independent in casePr(AB) = Pr(A)Pr(B)
A set of events {Ai} is independent in case
Example: Drug test
Pr( ) Pr( )i iii A A
Women Men
Success 200 1800
Failure 1800 200
A = {A patient is a Women}
B = {Drug fails}
Will event A be independent of event B ?
Copyright Tapan P Bagchi 48
Independence
Consider the experiment of tossing a coin twice Example I:
A = {HT, HH}, B = {HT} Will event A be independent from event B?
Example II: C = {HT}, D = {TH} Will event C be independent from event D?
Disjoint Independence
If A is independent from B, B is independent from C, will A be independent from C?
3/11/2012
25
Copyright Tapan P Bagchi 49
If A and B are events with Pr(A) > 0, the conditional probability of B given A is
Conditioning
Pr( )Pr( | )Pr( )
ABB AA
Copyright Tapan P Bagchi 50
If A and B are events with Pr(A) > 0, the conditional probability of B given A is
Example: Drug test
Conditioning
Pr( )Pr( | )Pr( )
ABB AA
Women Men
Success 200 1800
Failure 1800 200
A = {Patient is a Women}
B = {Drug fails}
Pr(B|A) = ?
Pr(A|B) = ?
3/11/2012
26
Copyright Tapan P Bagchi 51
If A and B are events with Pr(A) > 0, the conditional probability of B given A is
Example: Drug test
Given A is independent from B, what is the relationship between Pr(A|B) and Pr(A)?
Conditioning
Pr( )Pr( | )Pr( )
ABB AA
Drug’s performance
Women Men
Success 200 1800
Failure 1800 200
A = {Patient is a Women}
B = {Drug fails}
Pr(B|A) = ?
Pr(A|B) = ?
Copyright Tapan P Bagchi 52
Bayes’ Rule for computing Conditional
Probabilities
Bayes’ Rule allows you to go from P(A/B) to P(B/A)
3/11/2012
27
Copyright Tapan P Bagchi 53
Cancer Test example A medical test is used to check cancer.
This test has a known reliability:P(Test +ive /person has cancer) = 0.92P(Test +ive /person healthy) = 0.04
We know that cancer is rare and in the general population P(cancer) = 0.001 = (0.1%)
If a person is randomly selected and his test is +ive, what is the chance that he has cancer?
Copyright Tapan P Bagchi 54
Data for the medical testP(cancer) = P(c) = 0.001, P(healthy) = 0.999P(test +ive/ c) = 0.92, P(test +ive / healthy) = 0.04
Question: Will you rely on this test to start a treatment for cancer?
3/11/2012
28
Copyright Tapan P Bagchi 55
Given two events A and B and suppose that Pr(A) > 0. Then
Bayes’ Rule for finding P(B/A)
)Pr()Pr()|Pr(
)Pr()Pr()|Pr(
ABBA
AABAB
Bayes Definitions:
Pr(B) = a priori
Pr(B|A) = a posteriori
Copyright Tapan P Bagchi 56
Bayes’ Rule
Pr(W|R) R R
W 0.7 0.4
W 0.3 0.6
Events:
R: It rains
W: The grass is wet
R W
Information
Pr(W|R)
Inference
Pr(R|W)
3/11/2012
29
Copyright Tapan P Bagchi 57
Bayes’ Rule
Pr( | ) Pr( )Pr( | )Pr( )
E H HH EE
Hypothesis H Evidence EInformation: Pr(E|H)
Inference: Pr(H|E)
PriorLikelihoodPosterior
Note: Pr(H)
= Pr(H is true)
Copyright Tapan P Bagchi 58
Solution to the medical testP(cancer) = P(c) = 0.001, P(healthy) = 0.999P(test +ive/ c) = 0.92, P(test +ive / healthy) = 0.04
P(+ive / c ) P(c)P( c / test +ive) = ---------------------------------------------------------
P(+ive/ c) P(c) + P(+ive/healthy) P(healthy)
Verify that the answer is 0.0225
Managerial question:Will you rely on this test to get a treatment?
3/11/2012
30
Copyright Tapan P Bagchi 59
Bayes’ Rule: More Complicated contd. Suppose that B1, B2, … Bk form a partition of S:
Suppose that Pr(Bi) > 0 and Pr(A) > 0. Then
; i j iiB B B S
1
1
Pr( | ) Pr( )Pr( | )Pr( )
Pr( | ) Pr( )
Pr( )
Pr( | ) Pr( )
Pr( ) Pr( | )
i ii
i ik
jj
i ik
j jj
A B BB AA
A B B
AB
A B B
B A B
Copyright Tapan P Bagchi 60
A More Complicated Example
R It rains
W The grass is wet due to rain or sprinkler
U People bring umbrella
Pr(UW|R)=Pr(U|R)Pr(W|R)Pr(UW| R)=Pr(U| R)Pr(W| R)
R
W U
Pr(W|R) R R
W 0.7 0.4
W 0.3 0.6
Pr(U|R) R R
U 0.9 0.2
U 0.1 0.8
Pr(U|W) = ?
Pr(R) = 0.8
3/11/2012
31
Copyright Tapan P Bagchi 61
A More Complicated Example
R It rains
W The grass is wet
U People bring umbrella
Pr(UW|R)=Pr(U|R)Pr(W|R)Pr(UW| R)=Pr(U| R)Pr(W| R)
R
W U
Pr(W|R) R R
W 0.7 0.4
W 0.3 0.6
Pr(U|R) R R
U 0.9 0.2
U 0.1 0.8
Q. What is the probability that people will bring umbrella when they see that grass is wet = Pr(U|W) = ?
Pr(R) = 0.8
Copyright Tapan P Bagchi 62
A More Complicated Example
R It rains
W The grass is wet
U People bring umbrella
Pr(UW|R)=Pr(U|R)Pr(W|R)Pr(UW| R)=Pr(U| R)Pr(W| R)
R
W U
Pr(W|R) R R
W 0.7 0.4
W 0.3 0.6
Pr(U|R) R R
U 0.9 0.2
U 0.1 0.8
Pr(U|W) = ?
Pr(R) = 0.8
3/11/2012
32
Copyright Tapan P Bagchi 63
Random variables and probability distributions
Copyright Tapan P Bagchi 64
Probability DistributionThe probability distributionrefers to the frequency with which all different possible outcomes occur. There are numerous types of probability distributions.
Small fish got caught first!
In CA if you fish in Fall, you are more likely to catch big fish
3/11/2012
33
Copyright Tapan P Bagchi 65
Random Variable and Distribution
A random variable X is a numerical outcome of a random experiment
The distribution of a random variable is the collection of possible outcomes along with their probabilities: Discrete case: Continuous case:
Pr( ) ( )X x p x
Pr( ) ( )b
aa X b p x dx
Copyright Tapan P Bagchi 66
Random Variable: Example Let S be the set of all sequences of three rolls of a
die. Let X be the sum of the number of dots on the three rolls.
What are the possible values for X? Pr(X = 5) = ?, Pr(X = 10) = ?
3/11/2012
34
Copyright Tapan P Bagchi 67
Expectation
A random variable X~Pr(X=x). Then, its expectation is
In an empirical sample, x1, x2,…, xN,
Continuous case:
Expectation of sum of random variables
[ ] Pr( )xE X x X x
11[ ] N
iiE X xN
[ ] ( )E X xp x dx
1 2 1 2[ ] [ ] [ ]E X X E X E X
Copyright Tapan P Bagchi 68
Expectation: Example Let S be the set of all sequence of three rolls of a
die. Let X be the sum of the number of dots on the three rolls.
What is E(X)?
Let S be the set of all sequence of three rolls of a die. Let X be the product of the number of dots on the three rolls.
What is E(X)?
3/11/2012
35
Copyright Tapan P Bagchi 69
Variance
The variance of a random variable X is the expectation of (X-E[x])2 :
Range = Max(Xi) – Min(Xi)
2
2 2
2 2
2 2
( ) (( [ ]) )
( [ ] 2 [ ])
( [ ] )
[ ] [ ]
Var X E X E X
E X E X XE X
E X E X
E X E X
Copyright Tapan P Bagchi 70
Finding probabilities by Counting Events
Many times probabilities can be determined by enumerating (listing all possible) events or counting the ones of our interest.
Examples: Tossing a perfect coin Getting exactly 4 heads in 10 tosses
3/11/2012
36
Copyright Tapan P Bagchi 71
PermutationsThe number of possible permutations (sequences) is the number of different orders in which particular events occur. The number of possible permutations are
where r is the number of events in the series, n is the number of possible events, and n! denotes the factorial of n = the product of all the positive integers from 1 to n.
N p n!
(n r )!
Copyright Tapan P Bagchi 72
CombinationsWhen the order or sequence in which the events occurred is of no interest, we are dealing with combinations. The number of possible combinations is
where r is the number of events in the series, n is the number of possible events, and n! denotes the factorial of n = the product of all the positive integers from 1 to n.
Nc n
r
n!r!(n r)!
3/11/2012
37
Copyright Tapan P Bagchi 73
The uniform distribution
A random variable is said to be uniformly distributed if the probability of all possible outcomes are equal to one another. Thus, the probability P(i), where i is one of n possible outcomes, is
P(i) 1n
12…i n X X is discrete here
1/n
Copyright Tapan P Bagchi 74
Bernoulli Distribution
The outcome of an experiment can either be success (i.e., 1) and failure (i.e., 0).
Pr(X=1) = p, Pr(X=0) = 1-p, or
E[X] = p, Var(X) = p(1-p)
1( ) (1 )x xp x p p
3/11/2012
38
Copyright Tapan P Bagchi 75
The binomial distribution
The mean and variance of a binomially distributed variable are given by
np
npqVariance Example:
P(x) = Prob of finding x heads in n tosses of a coin
Application: Acceptance Sampling in Quality Control—d defectives in a sample of size n.
Copyright Tapan P Bagchi 76
The binomial distribution
A process that has only two possible outcomes is called a binomial process. In statistics, the two outcomes are frequently denoted as success and failure. The probabilities of a success or a failure are denoted by p and q, respectively. Note that p + q = 1. The binomial distribution gives the probability of exactly k successes in n trials
P(k) n
k
pk 1 p n k
3/11/2012
39
Copyright Tapan P Bagchi 77
Binomial Distribution n draws of a Bernoulli distribution Xi~Bernoulli(p), X=i=1
n Xi, X~Bin(p, n) Random variable X stands for the number of times
that experiments are successful.
E[X] = np, Var(X) = np(1-p)
(1 ) 1,2,...,Pr( ) ( )
0 otherwise
x n xnp p x n
X x p x x
Copyright Tapan P Bagchi 78
Plots of Binomial Distribution
3/11/2012
40
Copyright Tapan P Bagchi 79
The Poisson distribution
Siméon Denis Poisson1781-1840
Siméon Denis Poisson1781-1840
Copyright Tapan P Bagchi 80
The Poisson distributionWhen the probability of “success” is very small, e.g., the probability of a gene mutation, then pk and (1 – p)n – k
become too small to calculate exactly by the binomial distribution. In such cases, the Poisson distributionbecomes useful. Let l be the expected number of successes in a process consisting of n trials, i.e., l = np. The probability of observing k successes is
Mean and variance are given by m = l and V = l, respectively.
P ( k ) k e
k!
3/11/2012
41
Copyright Tapan P Bagchi 81
Examples of Poisson Distribution
Number of buses arriving at a bus stop/hour Number of road accidents/week on NH6 Number of copying machine
breakdowns/month Number of customers arriving/hour at BC Roy
Hospital Applications: Service facilities design
(queuing theory, Communications networks), sampling plan design (quality control)
Copyright Tapan P Bagchi 82
Poisson Distribution Coming from Binomial distribution Fix the expectation =np Let the number of trials nA Binomial distribution will become a Poisson distribution
E[X] = , Var(X) =
otherwise0
0!)()Pr( xe
xxpxXx
3/11/2012
42
Copyright Tapan P Bagchi 83
Plots of Poisson Distribution
Copyright Tapan P Bagchi 84
Normal (Gaussian) Distribution X~N(,)
E[X]= , Var(X)= 2
If X1~N(1,1) and X2~N(2,2), X= X1+ X2 ?
2
22
2
22
1 ( )( ) exp22
1 ( )Pr( ) ( ) exp22
b b
a a
xp x
xa X b p x dx dx
3/11/2012
43
Copyright Tapan P Bagchi 85
Other important Distributions
Normal distribution N(m, s2) Control charts in QC
Weibull distribution Reliability
Chi Square distribution Market research
F distribution Six SigmaBeta distribution and the Triangular distribution
Project Mgmt
Poisson distribution Traffic studies
Copyright Tapan P Bagchi 86
Statistical Inference
Consider a coin. To determine if this is a fair coin, you flip it 10 times, with 3 heads and 7 tails. Do you think this is fair coin?
In other words, is = P[head] = 0.5?
3/11/2012
44
Copyright Tapan P Bagchi 87
Statistical Inference inferring parameters from observed data
Problem:
Likelihood function
This is really a joint probability of {Yi} calculated using Approach: Maximum likelihood estimation (MLE), or
maximize log-likelihood
Copyright Tapan P Bagchi 88
Example: Flip Coins to find
3/11/2012
45
Copyright Tapan P Bagchi 89
Example: Flip Coins (cont’d)Likelihood = joint probability of the observed data /
Maximize ln()
Copyright Tapan P Bagchi 90
Recall Probability: Basic IdeasTerminology: Trial: each time you repeat an experiment Outcome: result of an experiment Random experiment: one with random
outcomes (cannot be predicted exactly) Relative frequency: how many times a
specific outcome occurs within the entire experiment.
3/11/2012
46
Copyright Tapan P Bagchi 91
Statistics: Basic Ideas Statistics is the area of science that deals with
collection, organization, analysis, and interpretation of data.
It also deals with methods and techniques that can be used to draw conclusions about the characteristics of a large number of data points--commonly called a population--
By using a smaller subset of the entire data.
Copyright Tapan P Bagchi 92
For Example… You work in a cell phone factory and are asked to
remove cell phones at random off of the assembly line and turn it on and off.
Each time you remove a cell phone and turn it on and off, you are conducting a random experiment.
Each time you pick up a phone is a trial and the result is called an outcome.
If you check 200 phones, and you find 5 bad phones, then
relative frequency of failure = 5/200 = 0.025
3/11/2012
47
Copyright Tapan P Bagchi 93
Statistics in Quality Control
As engineers perform experiments, they collect data that can be used to explain relationships better and to reveal information about the quality of products and services they provide.
Copyright Tapan P Bagchi 94
Frequency Distribution: Scores for an Six Sigma class are as follows: 58, 95, 80, 75, 68, 97,
60, 85, 75, 88, 90, 78, 62, 83, 73, 70, 70, 85, 65, 75, 53, 62, 56, 72, 79
To better assess the success of the class, we make a frequency chart:
3/11/2012
48
Copyright Tapan P Bagchi 95
Now the information can be better analyzed. For example, 3 students did poorly, and 3 did
exceptionally well. We know that 9 students were in the average range of 70-79. We can also show this data in a freq. histogram (PDF).
Divide each no. by 26
Copyright Tapan P Bagchi 96
Cumulative Frequency The data can be further organized by calculating the
cumulative frequency (CDF). The cumulative frequency shows the cumulative number of
students with scores up to and including those in the given range. Usually we normalize the data - divide 26.
3/11/2012
49
Copyright Tapan P Bagchi 97
Measures of Central Tendency & Variation Systematic errors, also called fixed errors, are errors
associated with using an inaccurate instrument. These errors can be detected and avoided by properly
calibrating instruments Random errors are generated by a number of
unpredictable variations in a given measurement situation. Mechanical vibrations of instruments or variations in line
voltage friction or humidity could lead to random fluctuations in observations.
Copyright Tapan P Bagchi 98
When analyzing data, the mean alone cannot signal possible mistakes. There are a number of ways to define the dispersion or spread of data.
You can compute how much each number deviates from the mean, add up all the deviations, and then take their average as shown in the table below.
3/11/2012
50
Copyright Tapan P Bagchi 99
As exemplified in Table 19.4, the sum of deviations from the mean for any given sample is always zero. This can be verified by considering the following:
Where xi represents data points, x is the average, n is the number of data points, and d, represents the deviation from the average.
x 1n
x ii1
n
di (x i x )
Copyright Tapan P Bagchi 100
Therefore the average of the deviations from the mean of the data set cannot be used to measure the spread of a given data set.
Instead we calculate the average of the absolute values of deviations. (This is shown in the third column of table 19.4 in your textbook)
For group A the mean deviation is 290, and Group B is 820.
We can conclude that Group B is more scattered than A.
dii1
n
x ii1
n
x i1
n
dii1
n
nx nx 0
3/11/2012
51
Copyright Tapan P Bagchi 101
Variance Another way of measuring the data is by
calculating the variance. Instead of taking the absolute values of each
deviation, you can just square the deviation and find the means.
(n-1) makes estimate unbiased
v i1
n
(xi x )2
n 1
Copyright Tapan P Bagchi 102
Taking the square root of the variance which results in the standard deviation.
The standard deviation can also provide information about the relative spread of a data set.
Range (Xmax – Xmin) can also show spread
s i1
n
(x i x )2
n 1
3/11/2012
52
Copyright Tapan P Bagchi 103
The mean for a grouped distribution is calculated from:
wherex = midpoints of a given rangef = frequency of occurrence of data in the
rangen = f = total number of data points
x (xf )n
Copyright Tapan P Bagchi 104
The standard deviation for a grouped distribution is calculated from:
s (x x )2 f
n 1
3/11/2012
53
Copyright Tapan P Bagchi 105
Normal Distribution We could use the probability distribution from the figures
below to predict what might happen in the future. (i.e. next year’s students’ performance)
Copyright Tapan P Bagchi 106
Normal Distribution
Any probability distribution with a bell-shaped curve is called a normal distribution.
The detailed shape of a normal distribution curve is determined by its mean and standard deviation values.
3/11/2012
54
Copyright Tapan P Bagchi 107
Copyright Tapan P Bagchi 108
3/11/2012
55
Copyright Tapan P Bagchi 109
THE NORMAL CURVE
Using Table 19.11, approx. 68% of the data will fall in the interval of -s to s, one std deviation
~ 95% of the data falls between -2s to 2s, and approx all of the data points lie between -3s to 3s
For a standard normal distribution, 68% of the data fall in the interval of z = -1 to z = 1.
zi = (xi - x) / s
Copyright Tapan P Bagchi 110
AREAS UNDER THE NORMAL CURVE
z = -2 and z = 2 (two standard deviations below and above the mean) each represent 0.4772 of the total area under the curve.
99.7% or almost all of the data points lie between -3s and 3s.
3/11/2012
56
Copyright Tapan P Bagchi 111
Analysis of Two Histograms
Graph A is class distribution of numbers 1-10Graph B is class distribution of semester credits
Data for A = 5.64 +/- 2.6 (much greater spread than B)Data for B = 15.7 +/- 1.96 (smaller spread)Skew of A = -0.16 and Skew B = 0.146CV of A = 0.461 and CV of B = 0.125 (CV = SD/Mean)
Frequency A
01234567
2 3 4 5 6 7 8 9 10
Frequency B
0123456789
12 13 14 15 16 17 18 19 20
Copyright Tapan P Bagchi 112
References
1. Amir D Aczel and Jayavel Sounderpandian (2009). Complete Business Statistics, 6th
ed, Tata McGraw-Hill2. Bharat Jhunjhunwala (2008). Business
Statistics—A Self Study Textbook, 1st ed, S Chand
Recommended