03_Decision Trees and Expected Utility Flashcards
1
Q
Characterization of Decision Trees
A
- risk
- one criterion
- single DM
- Dynamic Decision
2
Q
What constitutes Dynamic Decision?
A
- DM has to make series of decisions
- between Decisions, a scenario occurs with a certain probability
3
Q
Decision Trees are generalizations of
A
Decision Matrices (static decision)
4
Q
Notation of a Decision Tree
- Decision Node
- Chance Node
- Result Node
- Edges
A
- Decision Node: Square
- Chance Node: Circle
- Result Node: **Triangle
- Edges: Line
5
Q
How do you solve decision tree?
A
roll-back procedure
- using the expected value or expected utility to evaluate different actions
6
Q
Important concepts in the context of Decision Trees
3 items
A
- value of test market
- value of perfect information
- sensitivity analysis
7
Q
How to Solve Decision tree with Expected Utility
Expected Utility Approach
A
- move all values from edges to result nodes and calculate final result
- determine u(x) based on best and worst outcome e+ and e-
- Transform outcomes e into utilities u(e)
- Calculate expected utility using roll-back procedure
8
Q
Scoring Model
Characterization of the Situation
A
- Deterministic
- Multiple Criteria
- Single DM
- Static Decision
9
Q
Idea of Scoring Model
A
- Modify Decision Matrix by introducing goals and corresponding weights instead of scenarios with probabilities
- solution approach however very **similar to decision making under risk **
10
Q
Procedure of Scoring Model
A
- sum over all weights w(j) = 1
- weights are determined according to importance of different goals
- Quantitative data transformed via value function with best outcome vj(ej+) = 1and wors outcome cj(ej-) = 0
- Transfomration of qualitative attributes to quantitative ones
- Evaluation via score of each alternative wj * vJ(ei,j)
11
Q
When do µ-sigma and Expected Utility Theory yield the same result?
3 items
A
- DM is risk neutral
- DM has quadratic utility function and parameters are calibrated accordingly
- results are normally distributed
12
Q
Scoring Model
6 Steps
A
- Determine objectives / goals
- Determine and normalize weights of objectives
- determine outcome values
- determine utikity function for every objective
- determine EU of every alternative
- select alternative with highest EU
