Examples of MCA

1
Examples of MCA

• Dating
• Best place to live or retire

2
(No Transcript)
3
(No Transcript)
4
Recent MESM projects with MCA

• Prioritize invasive species and treatment
  locations in Santa Monica Mtns.
• Prioritize sites
• Ventura Foothills, Blue oak woodland, Santa Clara
  River, Goleta Brownfields, Valley oak restoration
• Map stress index of watersheds
• Chiapas, Los Padres
• Campus Climate Neutral 1 policy ranking

5
What is Multicriteria Analysis?

• Decision analysis
• A set of systematic procedures for analyzing
  complex decision problems (multiple, conflicting
  and incommensurate objectives)
• The purpose of any GIS-based decision analysis
  is to provide insights and understanding, rather
  than to prescribe a correct solution. Often
  the process of attempting to structure the
  decision problem is more useful in achieving
  these aims than the numeric output of the
  GIS-based modeling. (Malczewski 2000)

6
Types of decisions

• Sites
• Pick best alternative (site)
• IHOP, landfill
• Identify set of good alternatives
• Top 10 beaches or destinations
• Set of graduate schools to apply to
• Rank all alternatives (i.e., a map)
• Regional vulnerability, sustainability index
• Set of Sites
• Pick best region or area
• Areas for agriculture, corridors

7
Elements of multicriteria analysis

• Goal(s)
• Decision maker(s)
• Preferences (weights)
• Attitude toward risk
• Evaluation criteria
• Objectives (desired state)
• Attributes (measure performance in relation to
  objectives)
• Alternatives
• Outcomes of alternative by criteria
• Decision rules
• Sensitivity analysis

8
Decision matrix
Source Malczewski 1999
Rank 1
Rank 2
Rank m
9
Hierarchy of criteria

• Many to many relationship possible

Source Malczewski 1999
10
Criterion map scales
Source Malczewski 1999
11
Point allocation
Source McCoy et al. 2003
12
Why weight?

• To express the importance (to the decision maker)
  of each criterion in relation to each other
• But also dependent on the range of criteria
  values
• Why not just ask decision maker for their
  weights?
• What if decision maker cannot weight criteria?

13
AHP pairwise comparison

• Calculate weights and consistency ratio from
  comparison matrix

14
Simplified version of AHP
Source Strager and Rosenberger 2005
15
Goal Identify high priority lands for protecting
terrestrial biodiversity in California
Source Regan et al. in press
16
Basketball analogy of risk

• Goal assemble a championship caliber team for a
  given budget
• Alternatives
• Kobe Bryant with four mediocre players
• Five good (but not great) players
• What if Kobe got hurt?
• What if the other four players (or even one of
  them) was terrible?
• In an environmental context
• map analysis was based on erroneous data or
  things change
• weakest link

17
Choice of decision rule How to aggregate
criteria?
Source Moffett and Sarkar 2006
18
Integration matrix
Source WWF
19
Boolean logic

• AND logic
• OR logic

Source Jiang and Eastman 2000
20
Weighted Linear Combination

• Compensatory tends to average
• Assumptions
• Linearity desirability of additional attribute
  unit is constant for any level of attribute
• Additivity no interaction (correlation) between
  attributes
• Tends to be ad hoc with little theoretical support

21
Ideal point or compromise programming

• Orders a set of alternatives on the basis of
  their distance from an ideal point in
  multicriteria space
• Can also consider the maximum distance from
  negative ideal (risk-averse)
• Compromise method prefers closest to ideal and
  farthest from negative ideal

22
Concordance/Discordance Analysis (Outranking)

• Based on pairwise comparison of alternatives
• Concordance is all criteria for which A is not
  worse than B (but not how much better)
• Discordance is all criteria for which A is worse
  than B (but not how much better)

23
VisualizationQuantiles
24
VisualizationRadar plots
25
VisualizationConsumers Reports
26
Highlights

• When to use MCA instead of statistics or process
  models?
• Rich literature in multicriteria analysis
• Dont reinvent the wheel or make a wobbly wheel
• Start from objectives, not data
• Select method that is consistent with
• Assumptions (ranking outcomes, criteria, and
  alternatives, risk)
• Type of decision (best alternative, good
  alternatives, or rank all alternatives)
• Criteria are usually facts, weights are social
  values