JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale...
Transcript of JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale...
![Page 1: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/1.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
JDEP 384H: Numerical Methods in Business
Instructor: Thomas ShoresDepartment of Mathematics
Lecture 27, April 24, 2007110 Kaufmann Center
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 2: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/2.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
Outline
1 NT: Decision Analysis and Game TheoryAn Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
2 Odds and Ends
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 3: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/3.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Outline
1 NT: Decision Analysis and Game TheoryAn Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
2 Odds and Ends
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 4: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/4.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Outline
1 NT: Decision Analysis and Game TheoryAn Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
2 Odds and Ends
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 5: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/5.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Sidebar: Grades and Final
What's the Score?
The grade lines are
A: 87%, B: 76%, C: 60%I will use this for both homework and midterm.Final Exam scale will be similar.
And what about the �nal?
Will be all take-home. Check the course directory for lastyear's �nal exam key.Will be available on the web in our course directory on Friday,April 27, 1:00 pm.Will be due on Thursday, April 3, 1:00 pm.
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 6: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/6.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Sidebar: Grades and Final
What's the Score?
The grade lines are
A: 87%, B: 76%, C: 60%I will use this for both homework and midterm.Final Exam scale will be similar.
And what about the �nal?
Will be all take-home. Check the course directory for lastyear's �nal exam key.Will be available on the web in our course directory on Friday,April 27, 1:00 pm.Will be due on Thursday, April 3, 1:00 pm.
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 7: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/7.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Sidebar: Grades and Final
What's the Score?
The grade lines are
A: 87%, B: 76%, C: 60%I will use this for both homework and midterm.Final Exam scale will be similar.
And what about the �nal?
Will be all take-home. Check the course directory for lastyear's �nal exam key.Will be available on the web in our course directory on Friday,April 27, 1:00 pm.Will be due on Thursday, April 3, 1:00 pm.
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 8: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/8.jpg)
Review Model Problem Data
The Payo� Table and Conditional Probability Data:
Payo�sAlternatives
States of Nature
Acceptable Unacceptable
Develop IPSell IP
Prior Probabilities
$7M −$1M$0.85M $0.85M
0.25 0.75
Conditional ProbsRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
0.6 0.20.4 0.8
Cost of Consulting: $30,000 = $0.03M.
![Page 9: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/9.jpg)
Review Model Problem Data
The Payo� Table and Conditional Probability Data:
Payo�sAlternatives
States of Nature
Acceptable Unacceptable
Develop IPSell IP
Prior Probabilities
$7M −$1M$0.85M $0.85M
0.25 0.75
Conditional ProbsRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
0.6 0.20.4 0.8
Cost of Consulting: $30,000 = $0.03M.
![Page 10: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/10.jpg)
Review Model Problem Data
The Payo� Table and Conditional Probability Data:
Payo�sAlternatives
States of Nature
Acceptable Unacceptable
Develop IPSell IP
Prior Probabilities
$7M −$1M$0.85M $0.85M
0.25 0.75
Conditional ProbsRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
0.6 0.20.4 0.8
Cost of Consulting: $30,000 = $0.03M.
![Page 11: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/11.jpg)
Review Model Problem Data
The Payo� Table and Conditional Probability Data:
Payo�sAlternatives
States of Nature
Acceptable Unacceptable
Develop IPSell IP
Prior Probabilities
$7M −$1M$0.85M $0.85M
0.25 0.75
Conditional ProbsRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
0.6 0.20.4 0.8
Cost of Consulting: $30,000 = $0.03M.
![Page 12: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/12.jpg)
Review Model Problem Data
The Payo� Table and Conditional Probability Data:
Payo�sAlternatives
States of Nature
Acceptable Unacceptable
Develop IPSell IP
Prior Probabilities
$7M −$1M$0.85M $0.85M
0.25 0.75
Conditional ProbsRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
0.6 0.20.4 0.8
Cost of Consulting: $30,000 = $0.03M.
![Page 13: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/13.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
An Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
Outline
1 NT: Decision Analysis and Game TheoryAn Intelligent Opponent: Game TheoryAn Indi�erent Opponent: NatureDecision Making with Experimentation
2 Odds and Ends
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 14: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/14.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 15: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/15.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 16: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/16.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 17: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/17.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 18: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/18.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 19: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/19.jpg)
Probability Background
Table of Conditional Probabilities and what to use on them:
ConsultantRecommendations
States of Nature
Acceptable Unacceptable
DevelopSell
P (D |A) P (D |U)P (S |A) P (S |U)
Bayes' Decision Rule: Calculate the expected value of thepayo� for each alternative using the best available estimates ofthe probabilities of the states of nature.
Law of Total Probability: Given disjoint and exhaustiveevents E1,E2,, . . . ,En, and another event F ,P (F ) =
∑ni=1
P (F |Ei )P (Ej).
(Short) Bayes' Theorem: P (E | F ) ≡ P(F |E)P(E)P(F ) .
(Long) Bayes' Theorem: P (Ek | F ) ≡ P(F |Ek)P(Ek)Pni=1 P(F |Ei )P(Ej)
.
![Page 20: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/20.jpg)
Probability Connections
The table (or matrix Q for posterior probabilities) that we want:
State of Nature
Consultant Recommends
Develop Sell
AcceptableUnacceptable
P (A |D) P (A | S)P (U |D) P (U | S)
In matrix form Q can be calculated from Bayes' theorem as[P (A |D) P (A | S)P (U |D) P (U | S)
]=
[P(D |A)P(A)
P(D)P(S |A)P(A)
P(S)P(D |U)P(U)
P(D)P(S |U)P(U)
P(S)
]=[P (A) 00 P (U)
] [P (D |A) P (D |U)P (S |A) P (S |U)
]T [1
P(D) 0
0 1
P(S)
]and by the law of total probability unconditional probabilities
are
[P (D)P (S)
]=
[P (D |A) P (D |U)P (S |A) P (S |U)
] [P (A)P (U)
].
![Page 21: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/21.jpg)
Probability Connections
The table (or matrix Q for posterior probabilities) that we want:
State of Nature
Consultant Recommends
Develop Sell
AcceptableUnacceptable
P (A |D) P (A | S)P (U |D) P (U | S)
In matrix form Q can be calculated from Bayes' theorem as[P (A |D) P (A | S)P (U |D) P (U | S)
]=
[P(D |A)P(A)
P(D)P(S |A)P(A)
P(S)P(D |U)P(U)
P(D)P(S |U)P(U)
P(S)
]=[P (A) 00 P (U)
] [P (D |A) P (D |U)P (S |A) P (S |U)
]T [1
P(D) 0
0 1
P(S)
]and by the law of total probability unconditional probabilities
are
[P (D)P (S)
]=
[P (D |A) P (D |U)P (S |A) P (S |U)
] [P (A)P (U)
].
![Page 22: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/22.jpg)
Probability Connections
The table (or matrix Q for posterior probabilities) that we want:
State of Nature
Consultant Recommends
Develop Sell
AcceptableUnacceptable
P (A |D) P (A | S)P (U |D) P (U | S)
In matrix form Q can be calculated from Bayes' theorem as[P (A |D) P (A | S)P (U |D) P (U | S)
]=
[P(D |A)P(A)
P(D)P(S |A)P(A)
P(S)P(D |U)P(U)
P(D)P(S |U)P(U)
P(S)
]
=[P (A) 00 P (U)
] [P (D |A) P (D |U)P (S |A) P (S |U)
]T [1
P(D) 0
0 1
P(S)
]and by the law of total probability unconditional probabilities
are
[P (D)P (S)
]=
[P (D |A) P (D |U)P (S |A) P (S |U)
] [P (A)P (U)
].
![Page 23: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/23.jpg)
Probability Connections
The table (or matrix Q for posterior probabilities) that we want:
State of Nature
Consultant Recommends
Develop Sell
AcceptableUnacceptable
P (A |D) P (A | S)P (U |D) P (U | S)
In matrix form Q can be calculated from Bayes' theorem as[P (A |D) P (A | S)P (U |D) P (U | S)
]=
[P(D |A)P(A)
P(D)P(S |A)P(A)
P(S)P(D |U)P(U)
P(D)P(S |U)P(U)
P(S)
]=[P (A) 00 P (U)
] [P (D |A) P (D |U)P (S |A) P (S |U)
]T [1
P(D) 0
0 1
P(S)
]
and by the law of total probability unconditional probabilities
are
[P (D)P (S)
]=
[P (D |A) P (D |U)P (S |A) P (S |U)
] [P (A)P (U)
].
![Page 24: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/24.jpg)
Probability Connections
The table (or matrix Q for posterior probabilities) that we want:
State of Nature
Consultant Recommends
Develop Sell
AcceptableUnacceptable
P (A |D) P (A | S)P (U |D) P (U | S)
In matrix form Q can be calculated from Bayes' theorem as[P (A |D) P (A | S)P (U |D) P (U | S)
]=
[P(D |A)P(A)
P(D)P(S |A)P(A)
P(S)P(D |U)P(U)
P(D)P(S |U)P(U)
P(S)
]=[P (A) 00 P (U)
] [P (D |A) P (D |U)P (S |A) P (S |U)
]T [1
P(D) 0
0 1
P(S)
]and by the law of total probability unconditional probabilities
are
[P (D)P (S)
]=
[P (D |A) P (D |U)P (S |A) P (S |U)
] [P (A)P (U)
].
![Page 25: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/25.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 26: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/26.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 27: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/27.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 28: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/28.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 29: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/29.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 30: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/30.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 31: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/31.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 32: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/32.jpg)
Decision Trees
A Method for Visually Displaying the Problem and OrganizingComputational Work:
Draw a tree from left to right with square nodes for decisionsand round nodes for events.
Place decision and outcome labels above each branch, alongwith costs and expected payo�s below each branch and labelsand (best, prior or posterior) probabilities above each branch.
Fill in the tree by working from right to left.
For each decision node, compare expected payo�s of eachbranch and choose the decision with largest expected payo�, inaccordance with Bayes' Decision Rule.
Let's set this up at the board.
How can we use Matlab to help us out?
Let's de�ne a conditionals matrix and calculate theprobabilities of each recommendation and then.
![Page 33: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/33.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
Game Theory Model Problem
The Problem:
Dominant strategy elimination and the more generalmaximin/minimax strategies will not solve the following problem.
Strategy
Player 2
1 2 3
Player 1123
2 3 -2-1 4 03 -2 -1
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 34: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/34.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
Game Theory Model Problem
The Problem:
Dominant strategy elimination and the more generalmaximin/minimax strategies will not solve the following problem.
Strategy
Player 2
1 2 3
Player 1123
2 3 -2-1 4 03 -2 -1
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 35: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/35.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
A Sensitivity Analysis
The Original Model Problem:
Two companies compete for the bulk of a shared market for acertain product and plan to execute one of three strategies. Bothmarketing departments analyzed them and have arrived atessentially the same �gures for outcomes.
The three strategies are:
Better packaging.An advertising campaign.Slight price reduction.
Suppose there is considerable uncertainty about the payo� inthe case that both players make a slight reduction in price.How could we clearly illustrate the e�ect of changes in thepayo� on the weight that one of the companies puts on thisstrategy?
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 36: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/36.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
A Sensitivity Analysis
The Original Model Problem:
Two companies compete for the bulk of a shared market for acertain product and plan to execute one of three strategies. Bothmarketing departments analyzed them and have arrived atessentially the same �gures for outcomes.
The three strategies are:
Better packaging.An advertising campaign.Slight price reduction.
Suppose there is considerable uncertainty about the payo� inthe case that both players make a slight reduction in price.How could we clearly illustrate the e�ect of changes in thepayo� on the weight that one of the companies puts on thisstrategy?
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 37: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/37.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
A Sensitivity Analysis
The Original Model Problem:
Two companies compete for the bulk of a shared market for acertain product and plan to execute one of three strategies. Bothmarketing departments analyzed them and have arrived atessentially the same �gures for outcomes.
The three strategies are:
Better packaging.An advertising campaign.Slight price reduction.
Suppose there is considerable uncertainty about the payo� inthe case that both players make a slight reduction in price.How could we clearly illustrate the e�ect of changes in thepayo� on the weight that one of the companies puts on thisstrategy?
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business
![Page 38: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/38.jpg)
One More Decision Analysis
Problem:
A credit manager must decide whether to extend a $100,000 credit
to a new small business customer.
Customers of this type generally �t into categories of poorrisk, average risk or good risk, with a corresponding pro�t of−15%, 10% and 20%, respectively.
Prior experience indicates that 20% of customers in thiscategory are poor risk, 50% are average and 30% are good.
Also, the manager could consult a credit-rating company for a$2,500 fee.
With this record, should we consider them?
Strategy
Actual (in percent)
Poor Average Good
EvaluationPoor
AverageGood
50 40 2040 50 4010 10 40
![Page 39: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/39.jpg)
One More Decision Analysis
Problem:
A credit manager must decide whether to extend a $100,000 credit
to a new small business customer.
Customers of this type generally �t into categories of poorrisk, average risk or good risk, with a corresponding pro�t of−15%, 10% and 20%, respectively.
Prior experience indicates that 20% of customers in thiscategory are poor risk, 50% are average and 30% are good.
Also, the manager could consult a credit-rating company for a$2,500 fee.
With this record, should we consider them?
Strategy
Actual (in percent)
Poor Average Good
EvaluationPoor
AverageGood
50 40 2040 50 4010 10 40
![Page 40: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/40.jpg)
One More Decision Analysis
Problem:
A credit manager must decide whether to extend a $100,000 credit
to a new small business customer.
Customers of this type generally �t into categories of poorrisk, average risk or good risk, with a corresponding pro�t of−15%, 10% and 20%, respectively.
Prior experience indicates that 20% of customers in thiscategory are poor risk, 50% are average and 30% are good.
Also, the manager could consult a credit-rating company for a$2,500 fee.
With this record, should we consider them?
Strategy
Actual (in percent)
Poor Average Good
EvaluationPoor
AverageGood
50 40 2040 50 4010 10 40
![Page 41: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/41.jpg)
One More Decision Analysis
Problem:
A credit manager must decide whether to extend a $100,000 credit
to a new small business customer.
Customers of this type generally �t into categories of poorrisk, average risk or good risk, with a corresponding pro�t of−15%, 10% and 20%, respectively.
Prior experience indicates that 20% of customers in thiscategory are poor risk, 50% are average and 30% are good.
Also, the manager could consult a credit-rating company for a$2,500 fee.
With this record, should we consider them?
Strategy
Actual (in percent)
Poor Average Good
EvaluationPoor
AverageGood
50 40 2040 50 4010 10 40
![Page 42: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/42.jpg)
One More Decision Analysis
Problem:
A credit manager must decide whether to extend a $100,000 credit
to a new small business customer.
Customers of this type generally �t into categories of poorrisk, average risk or good risk, with a corresponding pro�t of−15%, 10% and 20%, respectively.
Prior experience indicates that 20% of customers in thiscategory are poor risk, 50% are average and 30% are good.
Also, the manager could consult a credit-rating company for a$2,500 fee.
With this record, should we consider them?
Strategy
Actual (in percent)
Poor Average Good
EvaluationPoor
AverageGood
50 40 2040 50 4010 10 40
![Page 43: JDEP 384H: Numerical Methods in Businesstshores1/Public/JDEP384hS07/Week16/J… · Final Exam scale will be similar. And what about the nal? Will be all take-home. Check the course](https://reader036.fdocuments.us/reader036/viewer/2022071219/6055647fbd5f562d8378c013/html5/thumbnails/43.jpg)
NT: Decision Analysis and Game TheoryOdds and Ends
Which Brings Us To...
THE END
Instructor: Thomas Shores Department of Mathematics JDEP 384H: Numerical Methods in Business