Decision Analysis Flashcards
Decision Analysis
It is a systematic, quantitative and visible approach to evaluate and adjust the important choices that the decision or the business faces. It uses a variety to tolls to evaluate all relevant information to add in decision-making process.
Cycle (4D’s)
- Discovery Stage: problem is recognised
- Design Stage: build a model
- Decision Stage: Analyse the model & interpret the results of the model
- Deployment Stage: we try to implement model. to see if we are seeing the improvement we want to see. This shows if the model is efficient or needs to be improved, starts back from the discovery stage.
Process of the Decision Analysis
- Assess the situation
- Construct a decision tree/table
- Analyse the decision tree/table
- Interpret results and draw inferences
- Implement solution
Decision Analysis
- Organises thinking about decision making
- Stimulates communication between decision makers
- Does not simply describe the situation
- Leads to insights which influence the decision maker
Used in
oil exploration, capacity planning, investment and product launch
Modelling
Simple models often leads to insight about complex problems.
Concentrate on
- objectives
- strategies
- scenarios.
What is the payoff for a strategy a possible scenario?
- Objective
Minimise cost Maximise sales Maximise market share Maximise quality Minimise times
Pay attention to the time horizon:
1 month, 1 year, 5 years
It will be affected by the time horizon
- Strategies
Strategy is the rule identifying a feasible choice for every decision that could be encountered
How to construct Strategies?
- Identify all points at which decision have to be made.
- Identify all the decisions that can be made at each point
- Construct strategies by considering all possible combinations of decision at each decision point. - 3 decision point with 2 decisions = 222= 8 strategies
Note that some decisions will restrict the decision points that can be reached in the future.
- Scenario
Scenario is the possible realisation of future uncertainty.
- Certain Enviromemt
- Uncertain Enviromemt
Answer:
What is likelihood of a scenario occurring?
We want to focus on making good decisions and in the long run the number of outcomes will be better.
How to construct Scenarios?
- Identify all uncertainties in the problem.
Such as a result of economic conditions, outcome of preventive repair. - Identify all the possible realisations of each uncertainty.
If it is successful or unsuccessful. - Construct scenarios by considering all possible combinations of the realisations of each uncertainty. All the possible realisations of the uncertainties
Note that some of these combinations may be impossible reducing the number of scenarios that need to be considered.
Format 1: Decision Tables
Present information in a structured compact format.
- List all the strategies to the left hand side
- List all scenarios above the table
- Insert Vi,j, the consequences of following strategy si when scenario oj occurs in column j or row i of the table.
The consequences can be cost or the reward of the problem
Advantage:
- easy to put on spreadsheet and find
Format 2: Decision Tree
A diagram that illustrates the correct ordering of actions and events in a decision analysis problem.
Highlights the sequential nature of both the decision making process and the manner in which the future is revealed.
Decision nodes (square) represent points at which decisions have to be taken
Outcome nodes(circle) represent points at which information about the future is revealed
Branches (lines) represent a possible sequence of decisions and outcomes
Advantage
- is a more efficient way to show the model
- it shows sequential of decision making
- visually easier to follow and identify the issues in the process.
Disadvantage
- may not be suitable for very large problems
- it can get very complicated to draw more and more decisions to make
Format 3: Influence Diagram
don’t need to draw in course
We can represent the consequence of the decision making and and is suitable for larger sized problems and doesn’t include as many details as the decision tree. Simplified as compared to decision tree.
Influence diagrams are less complex than decision trees and show relationships between features of the problem.
Decision nodes
represent points at which decisions have to be taken
Outcome nodes
represent points at which information about the future is revealed
Value nodes
represent points at which rewards are generated
Edges connecting one node to another indicate that the first node influences the second node
Disadvantage:
- lack of information
- payoff? decision outcomes?
Decision Criteria
For problems involving uncertainty, there is no general criterion for selecting a best strategy.
Examples of decision criteria include:
- maximin payoff,
- maximax payoff,
- Hurwicz criterion,
- minimax regret,
- maximum expected monetary value,
- maximum expected utility,
- maximum expected monetary value/minimum variance.
In uncertain, there is no best strategy
Different decision criteria depends on:
(clarify these things before starting)
- are appropriate for different attitudes to decision making;
- are appropriate for different problems;
- can lead to the selection of different strategies.
- Maximin payoff criterion
Selects a strategy to maximise the worst possible payoff.
- Pessimistic view of the future.
- Ignores the likelihood of scenarios.
- Best calculated from a decision table.
- Maximax payoff criterion
Selects a strategy to maximise the best possible payoff.
- Optimistic view of the future.
- Ignores the likelihood of scenarios.
- Best calculated from a decision table.
- Hurwicz criterion
Maximax payoff criterion is too optimistic and Maximin payoff criterion is too pessimistic.
Selects a strategy to maximise the weighted sum of the best possible payoff and the worst possible payoff.
It can balance the two extremes
- Minimax regret criterion
Regret is the difference between the payoff and the best possible payoff for the scenario that occurs.
The minimax regret criterion selects a strategy to minimise the maximum regret
- Decision maker is concerned about how the decision will look in hindsight
- Ignores the likelihood of scenarios.
- Best calculated from a regret matrix.
- Maximum expected monetary value (EMV) criterion
Selects a strategy to maximise the expected monetary value.
- Assumes decision maker is risk neutral.
- Considers the likelihood of scenarios.
Compute through Decision Table:
for each strategy, calculate the sum of the possible payoffs weighted by probabilities.
Compute through Decision tree:
- Grow the decision tree
- Assign probabilities to the event outcomes on the tree
- Assign the cash flows to the tree
- Fold back the decision tree; calculate the EMV at each outcome node; select the decision that leads to the best EMV at each decision node.
The optimal strategy consists of the optimal decision at each decision node.
Pitfalls of EMV
EMV assumes that the decision-maker is indifferent to the pattern of risk associated with the possible payoffs.
In some situations this makes EMV a poor decision criterion. - Look at example
Risk Profile
The risk profile for strategy is a list of all scenarios that can occur when the strategy is followed, along with the payoff and probabilities associated with those scenarios.
Might be some strategies have the same EMV, so the decision maker makes a choice based on the risk profile of these strategies.
A decision maker may prefer a strategy with less than the optimal EMV if its risk profile is significantly better than the optimal strategy.
- Trivial to extract from a decision table.
- Easily calculated from a decision tree.
Sometimes the alternative strategy has a better risk profile, as it has a lower probability to make loss. Sometimes people follow the optimal EMV strategy. In some cases the alternative strategy has a better risk profile and is a better strategy for the decision makers
