Data, Models, and Decisions I WrapUp Notes

1
Data, Models, and Decisions I Wrap-Up Notes

• Note These are the instructors notes for RPAD504
  designed to provide a high level overview of what
  we have tried to accomplish in this course, we
  hope you find these notes useful in organizing
  your thoughts for the final

2
Data Government Information Management

• Technical Skills
• One Table Data Bases
• Multi-Table Data Bases
• Economic, Organizational, and Political
  Dimensions to Information Management
• Forensic Mental Health Case
• VOODS Case Study
• Government Information Management as a Tool for
  Change
• Where to find out more

3
Models and Decisions Underlying Mathematical
Principles(only four big ones)

• Linear mathematics
• YBOBIXIB2X2...BNXN
• Matrices
•  Probability
• Definition of simple probability Pr (Event A)
  ... (objective and subjective)
• Joint Probabilities--Pr (Event A and Event B)
• Conditional Probabilities--Pr (Event A/ Event B)

• Dynamics
• Difference Equations--X(t)function of (X(t- 1),
  other parameters, other Y(t- 1))
• Markov chains as special case with state
  transition matrix
• Stock-and-flow as special form
•  Optimality

4
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Difference Equations
• Mathematical Form Dynamic/NonOptimal/Deterministi
  c/Usually Linear
• Global Assumptions
• Time "jumps" from one period to the next (no
  within period dynamics)
• Fixed set of relationships will hold from time
  zero to time final
• Variables usually aggregate, not detailed
  individual characteristics
• No explicit statements about optimal or best
  alternative possible
• The future is determined (not probabilistic) and
  can be described
• Predictable changes from one period to next is
  the key focus

5
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Markov Chains
• Mathematical Form Dynamic /Non-Optimal/
  Probabilistic?? /Linear
• Global Assumptions
• Special Case of difference equations--see all
  above
• State Transition matrix exists and can be
  computed (very restrictive)
• Exhaustive and mutually exclusive states
• Sum of probabilities for a row is exactly one

6
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Stock and Flow Models
• Mathematical Form Dynamic /Non-Optimal/
  Deterministic /Usually linear in 504
• Global Assumptions
• Special Case of difference equations--see all
  above
• World is comprised of stocks and explicit flows
  that change them
• Reinforcing and balancing loops of circular
  causation exist and are important
• Where to find out more PAD 624

7
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Exploratory Optimization
• Mathematical Form Static/Optimal/Deterministic/Li
  near
• Global Assumptions
• An explicit linear objective function exists and
  can be written down
• Explicit alternatives involving choices among
  "decision variables"
• Constraints define all the possible choice points
• The "optimal" choice can be computed
• Where to find out more PAD 620

8
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Multi-Attribute Utility (MAU) Models
• Mathematical Form Static/Optimal/Deterministic/Li
  near
• Global Assumptions
• Analyst can specify in advance all of the
  alternatives being considered
• Alternatives can be evaluated using a finite
  number (usually less than 7) attributes
• Data (objective or judgmental) exists on all
  attributers and can be scaled
• Weights (absolute and relative) can be assigned
  to the various attributes
• A linear combination of attribute scales and
  weights is often good enough
• Last two assumptions can be relaxed for
  non-linear analyses

9
Models and Decisions Specific Techniques and
Approaches are tied to "Global Assumptions"

• Decision Analysis
• Mathematical Form Static (mostly)/Optimal/Probabi
  listic/Linear
• Global Assumptions
• Future is defined by alternating choice points
  and probabilistic chance points
• States of nature can be described by well-defined
  alternatives with knowable probabilities
• Pay-offs for each alternative can be known and
  quantified
• Expected value is often a useful organizing
  framework
• MIN-MAX or MAXI-MIN and other organizing
  frameworks possible
• Best choice given probabilities, states of
  nature, and organizing principle can be computed

10
Models and Decisions General Propositions about
Formal Models and Decision-Making

• All decisions are made by individuals, groups of
  individuals, organizations, or legal and
  political processes.
• Data and formal models support these types of
  decision-making--formal models do not replace
  other forms of decision-making nor should they be
  used as decision-making tools.
•  

11
Models and Decisions General Propositions about
Formal Models and Decision-Making

• Models are abstractions based on "hard" data.
  Assumptions (global and specific assumptions), as
  well as intuition and judgment (values', "best
  guesses" about suspected or unknown effects, and
  subjective estimates to "fill in" data gaps)
• Hence, in the public sector at least 5 very
  different types of "decision domains" exist and
  the exact role of data and formal models differs
  from one decision domain to the next. Decision
  domains include (note other classifications of
  domains are possible)

12
Models and Decisions General Propositions about
Formal Models and Decision-Making

• Personal Decision Support
• Group Decision Support
• Organizational Decision Support
• Legal Decision Support
• Political Decision Support

13
Models and Decisions Seven Big Ideas

• Global Assumptions
• Specific Assumptions
• Parameters and Data
• Model Validity
• Sensitivity Analysis
• Implementation
• Utility or Usefulness

14
Models and Decisions Seven Big Ideas1

• Global Assumptions
• All formal models evoke a basic mathematical form
  that imposes a number of "global assumptions" on
  the problem being investigated. The analyst must
  be sure that the set of global assumptions do not
  distort a problem too much or force a "round
  problem to fit into a square analytic hole".

15
Models and Decisions Seven Big Ideas2

• Specific Assumptions
•  In addition to global assumptions, every formal
  model has a number of assumptions that are
  specific to the problem and model at hand. For
  example a MAU model assumes that the attributes
  listed are exhaustive and correct an
  optimization problem assumes that the given
  objective function captures some real world
  features of the problem being solved. Specific
  assumptions have to be made explicit, discussed
  openly and made to "match" the problem as it
  presents itself.

16
Models and Decisions Seven Big Ideas3

• Parameters and Data
• The issue here is where to the numbers come from
  and how good are they?
• Some parameters come from hard data and
  represent real effects "out there" in the world
  being modeled.
• Some parameters are "best guess" of the manager
  (or management team or organization).
• Yet other numbers represent subjective values of
  some manager, management team, or organization
  (such as the weights in a MAU model). These types
  of data should never be confused and assumptions
  about the data should be made open to scrutiny.

17
Models and Decisions Seven Big Ideas4

• Model Validity
• This is a complicated and less well-understood
  topic than many would have us to believe.
• The basic issue is do the model and its various
  classes of assumptions and types of data "fit"
  the problem being studied.
• Since all models are abstractions, in some sense
  all models are not "valid" because they do not
  contain all the features of reality. So we are
  always talking about whether or not the model is
  good enough or valid enough for some purpose(s).
• There are lots of formal tests that can help
  measure "closeness" between model assumptions and
  real data

18
Models and Decisions Seven Big Ideas5

• Sensitivity Analysis
• The general issue here is do the recommendations
  and conclusions coming out of a formal model
  change if one or more of the input parameters
  changes (e.g., a relative weight in a MAU model)
  or if one of the assumptions changes (e.g.
  another constraint is added to an optimization
  problem).

19
Models and Decisions Seven Big Ideas6

• Implementation
• Unless a formal model is being used for personal
  decision support, some other group of individuals
  or unit of an organization will have to be
  persuaded that the work that went into a model is
  correct and that its conclusions are correct and
  useful. Getting others to be aware of the results
  of a piece of analysis (e.g. getting the
  commissioner to pay attention) is part of the
  overall problem of implementing a formal model.

20
Models and Decisions Seven Big Ideas7

• Utility or Usefulness
• This final concept is an over-arching concept
  that asks was it worth the effort to do this
  formal piece of analysis.
• Obviously, utility is related to where the
  numbers come from, whether the model is sensitive
  to its parameters and assumptions, whether the
  model has some degree of technical validity, and
  how well the model and its results have been
  implemented.
• However, the relationship between final utility
  and any of these other concepts is not always
  simple and straightforward.

21
The Role of Computing1

• Computing is Essential for all this work.
• Computing is a powerful for helping to think
  about and reengineer organizational processes
• Computing makes data analysis and modeling easier
  and more accessible to less technical
  professionals

22
The Role of Computing2

• Integrated packages such as MS-OFFICE are good
  application packages for present state-of-the art
  
• These packages will soon be obsolete
• Lessons concerning decisions and models should
  not become obsolete so quickly

23
(No Transcript)