Lecture 19/20 - AM - Part 2 Flashcards
What are decision mainly made of?
Emotional, instinctual reasons. (90-95%)
What does MCA stand for?
Multi Criteria Analysis
What are the steps for MCA?
1) Select decision criteria or indicators
2) Weightings for ea indicator (0-5)
3) Indicator value for ea option
4) Calculate the overall results (multiply weighing with value and sum)
Disadvantages of MCA
- weighting hard to decide
- inconsistencies in weighting
Advantages of MCA
- fast and dirty
- combine w/ others = powerful
What is risk based decision making?
Risk = prob x consequence
identify high risk areas that should be addressed
What is risk used for?
Balance the funding gap.
Powerful when combined w/ Monte Carlo
What is analytical hierarchy?
Pairwise comparison method.
Advantages of analytical hierarchy?
- Based on mathematical decision theory and is industry standard.
- Generates ratio data
Disadvantages of analytical hierarchy?
- Labour intensive
- Participants perceive the tool as a black box if they do not understand the mathematical theory
Steps to analytical hierarchy process? (AHP)
1) Define objective and decision options.
2) Structure elements in criteria, sub-criteria, alternatives.
3) Pairwise comparisons of elements in ea group.
4) Calc weighting and consistency ratio.
5) Evaluate alternatives according to weighting.
6) Evaluate benefit cost analysis (BCA).
7) Make decision
Pairwise comparison scale?
from 1/9 to 9. if comparing two one would get say 1/3 and the other gets 3. 1 would mean that they are equal.
What’s in the criteria comparison matrix?
This is step 3. it is a matrix with all the comparison scores. columns mirror the rows.
What is the normalized matrix?
A normalized version of the comparison matrix. All column will add to 1.
What is {W}?
The normalized principal Eigenvector of C matrix.
Why do a consistency check?
Make sure it makes sense. If A > B and B > C C cannot be greater than A.
How do we check for consistency?
- Mat multiplication. {Ws} = [C]{W}.
- {consis} = {Ws}.*{1/W}.
Find lambda = average({consis}).
Find CI = (lambda-n)/(n-1) - Look up RI
- Find consist ratio CR = CI/RI
- If CR = 0.1 = revaluate [C]
How are subcriteria weightings determined?
They are proportional to the main criteria. sub criteria = criteria * sub criteria
Benefit cost the ahp scheme?
idk
What are the limitations of AHP?
- We are naturally consistent so matrix not perfect, may want to do multiple times
- Get lazy with larger matrix comparisons
- Only an approximation of {W} and lambda.
What is the Delphi method?
- Less labour intensive replacement for the AHP method.
- Minimiser variance
- Based on, group judgements more valid than individual, structured group more accurate than unstructured.
- Requires experts.
How does Delphi method work?
- Rounds
- Questionnaires.
- Revise
Sunk cost
focus on future cost/benefits
Self serving bias
If we win it is out success, if we lose, it is someone else fault.
Sudo-diagnostic information?
- Information that you see that isn’t really relevant but you think it is.
- You see there is a correlation and imply there is causation.
Subtlety diagnostic information?
- information that is important but you don’t think it is.
Conformation Bias?
Once you have decided on something, new information has less of an impact.
Inside view?
Always optimistic
Outside view?
less optimistic but more realistic.
Think up down?
Aims - conclusions
Actions -Causes
Think bottom up?
Conclusions - Aims methods etc.
Causes - Auctions
