DNSC6221_Lecture2
Transcript of DNSC6221_Lecture2
-
7/29/2019 DNSC6221_Lecture2
1/38
DNSC 6221
JUD
Lecture Set 02Bumsoo Kim
05/29/2012
The George Washington University School of Business
Department of Decision Sciences
1
-
7/29/2019 DNSC6221_Lecture2
2/38
Introduction to Probability Brief history of probability
Probability interpretations
Classical approach
Relative frequency approach
Subjective approach
Definition of probability
Laws of probability (basic principles)
Conditional probability
Statistical independence and mutually exclusive events
Bayes Theorem
Probability trees
2
-
7/29/2019 DNSC6221_Lecture2
3/38
Brief History of Probability
The use of probability as a concept to measure uncertaintydates back to hundreds of years ago. Probability has
found applications in areas as diverse as medicine,gambling, weather forecasting and the law.
It is generally believed that the mathematical theory ofprobability was started by the French mathematician
Blaise Pascal (1623-1662) and Pierre Fermat (1601-1665)when they succeeded in deriving exact probabilities forcertain gambling problems involving dice.
3
-
7/29/2019 DNSC6221_Lecture2
4/38
Probability Interpretations1. Frequency Approach:
The probability of a particular outcome can be obtained if the
process is repeated a large number of times under similar
conditions.Ex: tossing a coin
2. Classical Approach:
The classical approach is a mathematical formula based
approach in which the possible outcomes are enumerated
and exact probabilities are calculated.
Ex: rolling two dice
4
-
7/29/2019 DNSC6221_Lecture2
5/38
Probability Interpretations The classical approach would not work in processes where
we cannot describe the possible outcomes and there is no
way to determine the probabilities mathematically.Ex: Whether or not a project will be successful
3. Subjective Approach:
According to the subjective, or personal, interpretation ofprobability, the probability that a person assigns to a possible
outcome of some process represents her own judgment of the
likelihood that the outcome will be obtained (degree of belief).
Ex: Expert opinions
5
-
7/29/2019 DNSC6221_Lecture2
6/38
Basic Principles Note that the calculus of probability applies equally well no
matter which interpretation one prefers.
Likelihood, chance
Only considerrandom experiments
Characteristics of random experiments
Can specify all possible outcomes.
Cannot predict a specific outcome with certainty.
6
-
7/29/2019 DNSC6221_Lecture2
7/38
Basic Principles Ex:Todays closing price change of a security relative to
yesterday.
Probability is the yardstick for measuring the uncertainty of these
outcomes.
7
-
7/29/2019 DNSC6221_Lecture2
8/38
Venn Diagram of Outcomes
- Tool to represent probabilities
- Used in set theory to show the mathematical orlogical relationship between different sets
Outcomes of a roll of a die
Sample space: {1, 2, 3, 4, 5 ,6}Event A: outcome is even: {2, 4, 6}
Event B: outcome is > 3: {4, 5, 6}
-
7/29/2019 DNSC6221_Lecture2
9/38
Basic PrinciplesExample:
An automobile dealer sells two brands of new cars. One, C, is primarilyAmerican in origin; the other, G, is primarily Japanese. The dealerperforms repair work under warranty for both brands. Each warrantyjob is classified according to the primary problem to be fixed. If there ismore than one problem in a given job, all problems are listedseparately. Records for the past year indicate the following numbers ofproblems:
Problem AreaEngine Transmission Exhaust Fit/Finish Other Total
Brand C 106 211 67 133 24 541
G 21 115 16 24 6 182Total 127 326 83 157 30 723
a. What is the probability that a randomlychosen problem comes frombrand C?P( randomly chosen problem comes from brand C) = .748
9
-
7/29/2019 DNSC6221_Lecture2
10/38
Basic Principles
A B
Venn diagram
10
Mutually exclusive events: Only one of many events can
occur at the same time
-
7/29/2019 DNSC6221_Lecture2
11/38
Basic PrinciplesExample:
Problem AreaEngine Transmission Exhaust Fit/Finish Other Total
Brand C 106 211 67 133 24 541G 21 115 16 24 6 182Total 127 326 83 157 30 723
b. Serious problems are those involving the engine or transmission.What is the probability that a randomly chosen problem is serious?
P (Serious problem) = P(E) + P(T)= 127/723 + 326/723= 453/723= .627
11
-
7/29/2019 DNSC6221_Lecture2
12/38
Addition Law of Probability General Addition Law
P(A or B) = P(A) + P(B) P(A and B)
AB (A and B) has been counted twice
Subtract once
Example:
What is the probability that a randomly chosen problem comes from Brand
C or is an engine problem?
C denotes "Brand C. P(C) = 541/723
E denotes Engine problem". P(E) =127/723
P(C or E) = P(C) + P(E) P(C and E)
= 541/723 + 127/723 - 106/723= 562/723 = .777 12
-
7/29/2019 DNSC6221_Lecture2
13/38
Complements Law
A A
13
A` is also used to represent the complement of A.
-
7/29/2019 DNSC6221_Lecture2
14/38
Complements Law Ex:Birthday Problem
- There are n people in a room. What is the probability that at least twoof them have the same birthday?
Assumptions: No birthdays fall on Feb 29;
All 365 days are equally likely for each person; and
Birthdays are independent
A = {at least two people have same birthday}
A` ={no two people have the same birthday}P (A` ) = (365)(364)(363)[365 (n-1)] / (365)n
= .29 for n=30 so P(A)= 0.71for n=50 P(A)=.97
14
-
7/29/2019 DNSC6221_Lecture2
15/38
Conditional vs Marginal
Probability Conditional Probability
Conditional probability of event A occurring given that event B hasoccurred, denoted by P(A|B), is:
P(A|B) = P(A and B) / P(B),provided P(B) > 0.
A B
Aand B
15
Marginal Probability: Probability of one variable taking a valueirrespective of others, i.e; P(A), P(B)
-
7/29/2019 DNSC6221_Lecture2
16/38
Conditional vs Marginal
ProbabilityExample:
Problem AreaEngine Transmission Exhaust Fit/Finish Other Total
Brand C 106 211 67 133 24 541
G 21 115 16 24 6 182Total 127 326 83 157 30 723
a. For the auto dealers data, what is the probability that a randomly chosen problem is an engineproblem, given that it comes from brand C?
P(Engine problem | Brand C) = P(E|C) = P(E and C)/P(C)= (106/723)/(541/723)
=.196where P(Brand C and Engine Problem) = P(C and E) = 106/723and P(Brand C) = P (C) = 541/723
Note: P (E|C) = .196 while P(E) = .176The occurrence of event C affects P(E)
16
-
7/29/2019 DNSC6221_Lecture2
17/38
Multiplication Rule Multiplication Rule
From the definition of conditional probability,
P(A and B) = P(B) P(A|B)
Obviously,P(A and B) = P(A) P(B|A)
Example: Randomly pick (without replacement) 2 cardsfrom a standard deck. Find probability of 2 hearts.
A = {1st
card is a heart}, B = {2nd
card is a heart}P (A and B) = P(A) P(B|A) = (13/52) (12/51)
The multiplication rule is useful in Probability Trees.
17
-
7/29/2019 DNSC6221_Lecture2
18/38
Statistical Independence If occurrence of one event does not affect the other, they are
independent events.
Mathematically,A and B are independent events if and only if
P(A|B) = P(A)
Example:
Are events Brand C and Engine Problem independent?
P(Engine Problem) = P(E) = 127/723 = .176
P(Engine Problem|Brand C) = P(E|C) = .196
Conclusion:The events Brand C and Engine Problemare notindependent because
P(E|C) P(E)
18
-
7/29/2019 DNSC6221_Lecture2
19/38
Statistical Independence Multiplication Rule for Independent Events
If A and B are independent, then
P(A and B) = P(A) P(B)
Reason: From Multiplication Rule,
P(A and B) = P(A|B) P(B)
From independence,P(A|B) = P(A)
P(A and B) = P(A) P(B)
19
-
7/29/2019 DNSC6221_Lecture2
20/38
Independent vs Mutually
Exclusive Events Coin flip example:
One flip: Heads or Tails are mutually exclusive eventsTwo flips: The outcome of each flip is independent
If A and B are mutually exclusive events, then A and B can't bothhappen at the same time. Mutually exclusive events have nobasic outcomes in common: P( A and B) = 0
If A and B are independent, then P(A and B) = P(A) P(B)
Mutually exclusive events are always dependent.
20
-
7/29/2019 DNSC6221_Lecture2
21/38
Some Useful Definitions
Union of EventsThe union of two events A and B is all basic
outcomes that are part of either A or B or both:A B = A or B
Intersection of Events (Joint Probability)
The intersection of two events A and B is all basicoutcomes that are part of both A and B:
A B = A and B
P( A and B) = P( A , B)
-
7/29/2019 DNSC6221_Lecture2
22/38
Formulas so far Definition: 0
-
7/29/2019 DNSC6221_Lecture2
23/38
ExampleExample :
Suppose that A and B are mutually exclusive events.
P(A) =0.4 and P(B) =0.3
Find P(A and B), P(A or B), P(not A and not B), P(A and not
B), P(A|B), P(A|not B) and P(B|not A)
23
-
7/29/2019 DNSC6221_Lecture2
24/38
Probability Tables
- Tool to compute and represent probabilities
- Ex: Bivariate Joint Probabilities
B1 B2 . . . Bk
A1 P(A1 B1) P(A1 B2) . . . P(A1 Bk)
A2
P(A2 B1) P(A2 B2) . . . P(A2 Bk)
. . . . .
Ah
P(Ah B1) P(Ah B2) . . . P(Ah Bk)
-
7/29/2019 DNSC6221_Lecture2
25/38
Probability Trees
Many problems require the successive use of several of the basicprinciples to obtain a solution. These problems can be solvedalgebraically, but it is helpful to be familiar with some tools thatcan help you keep the logic straight. One of these tools isprobability (or decision) trees.
Probability trees use the multiplication law of probabilities.
In a probability tree, the probability for a specific path is found byusing the multiplication rule.
25
-
7/29/2019 DNSC6221_Lecture2
26/38
Probability TreesExercise:
A purchasing dept. finds that 75% of its special orders are
received on time. Of those orders that are on time, 80% meetspecifications completely; of those orders that are late, 60%
meet them.
T = {Order is on time} M = {Meets specifications}
P(T) = .75 P(M|T) = .80 P (M| not T ) = .60
26
-
7/29/2019 DNSC6221_Lecture2
27/38
Probability Trees
P(T) = .75
P( ) = .25
T
P(M|T) = .80
P( |T) = .20
P(M| ) = .60
M
M
P(T and M) = .60
M
P( and M) = .15T
T
P( | ) = .40M
T
T
P(T and ) = .15
P( and ) = .10
27
-
7/29/2019 DNSC6221_Lecture2
28/38
Probability Trees Find the probability that an order is on time and meets
specifications.
P(T and M) = .60
Find the probability that an order meets specifications.
P(M) = P(M and T) + P(M and not T)
= .60 + .15= .75
28
-
7/29/2019 DNSC6221_Lecture2
29/38
Probability TreesExercise:
A manufacturer of snack crackers introduces several new
products each year. About 60% of the introductions are failures,30% are moderate successes, and 10% are major successes.
To try to improve the odds, the manufacturer tests new products
in a customer tasting panel. Of the failures, 50% receive a poor
rating in the panel, 30% a fair rating and 20% receive a good
rating, For the moderate successes, 20% receive a poor rating,40% a fair rating, and 40% a good rating. For major successes,
the percentages are 10% poor, 30% fair and 60% good.
29
-
7/29/2019 DNSC6221_Lecture2
30/38
Probability Trees Find the joint probability of a new product being a failure and
receiving a poor rating.
Construct a probability table of all possible joint probabilities ofnew product results and panel rating.
If a new product receives a good rating, what is the probabilitythat the product will be a failure?
Create a probability tree using the given information. Use thetree to find the probability that a new product will be a majorsuccess, given that it gets a poor rating.
30
-
7/29/2019 DNSC6221_Lecture2
31/38
Bayes Theorem Extension of Conditional Probability
Result generalizes to Mevents,A1,A2, ...,AM
31
1
1
1 1
1 1 2 2
|
|| |
P A BP A B
P B
P B A P A
P B A P A P B A P A
-
7/29/2019 DNSC6221_Lecture2
32/38
Bayes Theorem Importance: Update of prior beliefs
P(A1) priorbelief about probability ofA1
P(B|A1) probability relating information / signal to outcome(based on information about past relationship between signal andoutcome)
Bayes Theorem provides means to calculate posterior
probability P(A1 | B)
32
-
7/29/2019 DNSC6221_Lecture2
33/38
Example: Market Research at
Arrow Cosmetics Arrow Cosmetics is considering the introduction of a
new location
Possible outcomes:
Units sold = {600K, 300K, 200K}
= {High, Medium, Low}
Subjective Probabilities:
33
High 0.3
Medium 0.4
Low 0.3
-
7/29/2019 DNSC6221_Lecture2
34/38
Additional Information Arrow can hire Marble Research to perform market
study test for new formula
Possible evaluations: {Hit, Flop}
Past experience with Marbles research:
P (Hit | High) = 1.0
P (Hit | Medium) = 0.375
P (Hit | Low) = 0.166
34
-
7/29/2019 DNSC6221_Lecture2
35/38
Revising Probabilities Updating the a priori/marginal probability with additional
information (signal,test,symptom) to come up withrevised/posterior/conditional probability
Suppose that Marble declares that the new product willbe a hit
Should Arrow adjust its belief about the probability thatsales will be high? If so, how?
P(High) Test P(High | Hit)
P(Medium) Test P(Medium | Hit)
P(Low) Test P(Low | Hit)
35
-
7/29/2019 DNSC6221_Lecture2
36/38
Lets Make a Deal! Suppose you are on a game show, and you are given the choice
of three doors. Behind one door is a car, behind the others,goats.
You pick a door, say number 1, and the host, who knows what'sbehind the doors, opens a door which has a goat. Say this isdoor 3.
The host asks you:
"Do you want to pick door number 2?"
Is it to your advantage to switch your choice of doors?
http://www.youtube.com/watch?v=mhlc7peGlGg
http://www.youtube.com/watch?v=P9WFKmLK0dc&feature=fvw
36
http://www.youtube.com/watch?v=mhlc7peGlGghttp://www.youtube.com/watch?v=P9WFKmLK0dc&feature=fvwhttp://www.youtube.com/watch?v=P9WFKmLK0dc&feature=fvwhttp://www.youtube.com/watch?v=mhlc7peGlGg -
7/29/2019 DNSC6221_Lecture2
37/38
Summary Interpretations of probability (classical, frequency and subjective
approaches)
Rules and definitions of probability;The addition law for mutually exclusive events and non mutually
exclusive events, the complements law of probability.
Marginal, joint and conditional probabilities
Statistical Independence
Probability Trees
37
-
7/29/2019 DNSC6221_Lecture2
38/38
Lecture 3 Topics:- Random variables
- Probability distributions, joint and marginal distributions
- Expected value, standard deviation (or variance) of a random
variable
- Covariance & Correlation
Things to do:- Go over these notes & Check Ch. 4.1, 4.2 for clarifications
- Read Chapter 4.3-4.9 in the textbook
- Attempt Questions 3-4 in Assignment 1