Decision analysis Flashcards
(27 cards)
Can you predict future
You can’t predict the future, but you can be
smart about how you manage in the face of risk & uncertainty. have a good understanding of cost& benefits
Goal of decision analysis
to help individuals and businesses make good decisions using a structured approach to decision due to less info & uncertainty
Character. of Decision Problem
-A set of alternatives (or strategies) available to the decision maker to solve the problem.
-The criteria in a decision problem represent various factors that are important to the decision maker and are the basis for making decisions about the alternatives. This is the payoff, usually in dollar$.
-The States of Nature in a decision problem
correspond to future events that are not under the decision maker’s control.
Payoff tables
shows the payoff, profit or loss for the range of possible outcomes.
Decision Rules
if you know what the outcome will be, you can choose the correct alt.
Non probabilistic methods
These rules help to enhance our insight and
sharpen our intuition about decision problems so we can make more informed decisions:
→ Maximax (Optimistic criteria)
→ Maximin (Conservative criteria)
→ Minimax regret
Maximax decision rule
MAXimizing the MAXimum (best) case
select best(max) payoff in each row(alter.) then select the alt with largest payoff.
Red flags for maxmax payoff
no guarantee that the maximum payoff
will occur, so this strategy is high risk. higher reward with higher risk.
maximax criterion is most appropriate
when the decision maker can survive even the worst-case payoff. decision(vlookup)
MAXIMIN Decision Rule
this rule’s strategy is selecting the alternative
that maximizes the minimum payoff. low risk (pessimistic approach-get the best result in a worst case scenario) conservative criteria
MINIMAX regret decision rule
based on the concept of regret/opp. loss
regret we feel if u make wrong choice given state of nature (outcome)
-1st convert to regret matrix then summ. the possible opp.
Probabilistic Methods
-can be used if the states of nature are assigned prob. that rep. their likelihood occurrence
-problems occurred more than 1 estimate prob. from historical data
-Many decision problems represent one-time
decisions so data for estimating prob. are unlikely to exist. In these cases, prob. are often assigned subjectively.
Expected Monetary Value
selects the dec. alt. with largest payoff based on prob. EMV= prob. * payoff
calculated for each outcome
EV of Perfect Information
Probabilities do not tell us which state of nature will occur; they only indicate the likelihood of the various states of nature
Decision Trees
Sows decision problems in a graphical format
can be used for any problem but mostly MULTI STAGE & CONTINGENT
Decision Tree Conventions
composed of nodes(square, circles, triangle) & branches
-Decision node sq. rep choice
-Event node circ. rep State of Nat. (no control)
-Terminal node trig. rep completion
Decision Tree Branches
-Time goes left to right
branch leading into node (left ) occurred
branch leading out node (right) not yet
-branches leading out of sqr rep possible dec
-branches leading out of O prob. node rep. outcome of uncertain events.
Rolling back a decision tree
way to calculate EMV to identify best dec.
-sq. alt. with highest EMV is selected
-O EMV is calc.
Multi stage Decisions
where decisions must be made at several different stages of the process.
Most decisions we make lead to other decisions
Bayes Theorem & contingent Dec.
Uncertain outcomes-
-Prior Prob. prob. of an event b4 new info.
-Posterior prob. prob after new info comes
Conditional Prob.
P(AIB) = prob. that A occurs given that B is to occur
Contingent decisions be met
before a decision is made
Perfect vs, Sample Info
PI is not realistic, SI is both economical & feasible. allows us to make more precise
estimates of the probabilities.
Exp. Val. of SI
SI is often expensive to obtain
This is the maximum amount we should be willing to pay to obtain sample information
EVSI = (A+ COST) - B
