Multicriteria decision analysis

1
Multi-criteria decision analysis

• List objectives and construct value trees
• Analyze decisions involving multiple criteria by
  means of an additive value function and interpret
  the answers sensibly.
• You should be aware of some of the difficulties
  of multi-criteria decision analysis. The answer
  may be too convincing!
• More detail and reading in other handout

2
Some examples

• Clemen (1996), or Clemen and Reilley (2001)
• Bowhill (1994) appraisal of options for a
  business school
• Times university ranking
• Choosing suppliers for a business
• Choosing a hall of residence
• Risk and return in finance
• Company quality assessments
• Choosing IS projects
• My transport to work decision

3
What criteria are relevant to this decision?

• The CEO of a large automobile company is
  concerned about the design of a new version of a
  well-known model. There is a risk of fuel-led
  fire.
• Plan is to produce 40 000 000 cars.
• The probability of explosion is 1/100 000.
• Maximum compensation of 200 000 euros per death
• The cost of re-design is 9 euros per car
• What do you think he should do?
• (From a presentation by Marc Menestrel who got it
  from Mark Ashbar The Corporation)

4
MCDA Terminology

• Alternatives, options, courses of action,
  strategies are words for the things between
  which the decision maker is choosing
• Objectives, criteria (or evaluation criteria),
  values are used to compare alternatives and
  decide which are best
• Attributes are qualities of the alternatives
  useful for applying these
• All words used in the ordinary English sense
  use your dictionary

5
Choose a decision and

• Construct a value tree
• List the alternative courses of action

6
An example to show the approach

• Please refer to the other handout on MCDA
• Ill use my decision about transport to work as
  an example (TransportDecisionDraNov06.xls)

7
More on multi-criteria decision analysis

• Diagrams efficient frontiers and option
  profiles
• Sensitivity analysis
• Things to check
• Examples

8
Diagrams are useful

• Option profiles (ie a profile of each option)
• Scatter diagrams of costs vs benefits to show the
  efficient frontier
• See examples

9
Caution!

• Additive value functions very easy to construct
  but
• you must check that
• the value tree is complete and without overlaps
• the value scores and weights are meaningful, and
• the other assumptions are OK. See in particular
  Step 2 and the notes at the end of Step 5 on the
  handout.
• Always worth doing a sensitivity analysis (see
  notes)
• Dont forget different DMs may have different
  weights and value scores and even different
  criteria.
• Eg Jennys choice of university

10
Example to show judgmental dependence

• When choosing a holiday would you prefer
• A hot climate or a cold climate?
• an open-air or an indoor swimming pool?
• (from Goodwin and Wright, 1998, p. 38)
• Take care when the answer involves the words it
  depends on
• May be helpful to change the way the values are
  defined

11
Exercises 2 and 3 on the other handout
12
More examples

• Times university rankings
• May be used by students to make decisions but
  should check values and weights are appropriate
• Why are all criteria except teaching and research
  quality equally weighted. What matters to the
  student?
• Finance
• Criteria are risk (sd) and return (mean/EMV)
• Trade-off normally avoided by dominance
  (efficient frontier) arguments
• Many businesses use scoring systems for
  aggregating values eg Deming and Baldridge
  quality assessment systems

13
Group decision making

• Decisions made within organisations may have
  different stakeholders with different goals etc
• Means that mcda more complex
• Structured methods to clarify rationale for
  decisions