2. Decision Analysis Examples
Oil field and services valuation (Shell, Chevron,
Schlumberger, …)
Pharmaceutical R&D decisions (Eli Lilly, …)
Chemical companies (Du Pont, Kodak, …)
Banks (contingency planning for data centers,
…)
Assisted living facility (Australia: choose
among four strategic plans)
Allocation of public funds (Germany:
evaluation of projects submitted to Berlin
Government Senate Department for
Economics)
3. Evaluation of Projects
Define objectives, e.g.:
Cost
Number of children reached
Expected improvement in
Enrollment
Performance
Graduation rates
Evaluate how a project fulfills each objective.
Determine probabilities.
Perform sensitivity analysis.
4. Probability Assessment
Decision and risk analyses typically rely on
probability assessments based on human
judgment.
Judgments should be informed by data when
data is available, but need not be entirely
based on data.
Is the past indicative of the future?
With new events with no obvious data, are there
historical analogies?
In assessing probabilities from experts, you
need to:
Be clear about the questions you are asking and the
responses given,
Ask questions that people can reliably answer, and
Be aware of potential biases, both cognitive and
motivational.
5. Methods for Assessing Probabilities:
Direct Assessment
What is your probability that more than 500 children
between ages 7-8 will be enrolled in school under
Project X?
Use of a reference gamble/Betting odds
Many resist direct assessments out of fear of
not giving the "right probability.“
Assessing conditional probability assessments may be
easier and have been shown to be more reliable.
What is your probability that … given that …
6. Psychological Pitfalls
Anchoring
Status Quo
Sunk-Cost
Confirming Evidence
Framing
Overconfidence
Recallability
Base-Rate
Prudence
Seeing Patterns where none exist
7. Quantity 10th Quantile 50th Quantile 90th Quantile
1 Year end close of Dow Jones
for 1962
2 Population of Panama (May
2010)
3 Number of patents for
invention issued by the US
Patent and Trademark office
(2010)
4 Airline distance between
Austin and New Delhi
(miles)
5 Birthday of Coach Gail
Goestenkors
(month/day/year)
6 Height of Hoover Dam (feet)
7 Number of windows in the
Pentagon
8 Diameter of Earth’s moon
(miles)
9 Worldwide airplane accident
deaths on scheduled flights
during 2010
10 Release of Star Wars IV
(month/day/year)
8. Quantity Value
1 Year end close of Dow Jones
for 1962
652.1
2 Population of Panama (May
2010)
3,405,813
3 Number of patents for
invention issued by the US
Patent and Trademark office
(2010)
219,614
4 Airline distance between
Austin and New Delhi
(miles)
8347.67
5 Birthday of Coach Gail
Goestenkors
(month/day/year)
2/26/63
6 Height of Hoover Dam (feet) 726.4
7 Number of windows in the
Pentagon
7,754
8 Diameter of Earth’s moon
(miles)
2,159
9 Worldwide airplane accident
deaths on scheduled flights
during 2010
817
10 Release of Star Wars IV
(month/day/year)
5/25/77
9. Which of the two programs would
you favor?
Imagine that the US is preparing for the
outbreak of an unusual Asian disease, which is
expected to kill 600 people. Two alternative
programs to combat the disease have been
proposed. Assume that the exact scientific
estimates of the consequences of the programs
are as follows:
If Program A is adopted, 200 people will be saved.
If Program B is adopted, there is 1/3 probability that
600 people will be saved, and 2/3 probability that no
people will be saved.
10. Which of the two programs would
you favor?
Imagine that the US is preparing for the
outbreak of an unusual Asian disease, which is
expected to kill 600 people. Two alternative
programs to combat the disease have been
proposed. Assume that the exact scientific
estimates of the consequences of the programs
are as follows:
If Program C is adopted, 400 people will die.
If Program D is adopted, there is 1/3 probability that
nobody will be die, and 2/3 probability that 600 people
will die.
11. Assigning Probabilities
Consult existing information
Collect new data
Ask experts
Unless they have statistical experience and
relevant data, most people cannot reliably
provide expected values or standard deviations.
Instead assess 5th, 50th and 95th percentiles.
Break uncertainties into new components
12. Using Assessed Percentiles in
Decision Trees
Say your expert provides the following percentiles
on the number of new enrollments from age
group 7-10 under project X:
5th percentile : 60 students
50th percentile : 120 students
95th percentile : 185 students
This assessment corresponds to the following
scenarios and probabilities: