Chapter 3 - Decision Analysis

Thompson Lumber Data Profit Probability Large Plant Small Plant Do nothing

Result Favorable Unfavorable Market Market 0.5 0.5 200,000 (180,000) 100,000 (20,000) -

EMV

Minimum

Maximum Coefficient

Maximum

10,000 40,000 40,000

(180,000) (20,000) -

200,000 100,000 200,000

100000 <- Expected value under certainty 40000 <- Best expected value 60000 <- Expected value of perfect information

Opportunity Loss Expected Value of Perfect Information Column best 200,000 -

Table of Regret (regret adalah nilai di mana harapan tidak sesuai dengan kenyataan; '0' adalah Tidak Menyesal)

Probability Large Plant Small Plant Do nothing

Favorable Unfavorable Market Market 0.5 0.5 180,000 100,000 20,000 200,000 Minimum

Expected Maximum 90,000 180,000 60,000 100,000 100,000 200,000 60,000 100,000

Keputusan: 1 Dengan model Maximax, buat Large Plant 2 Dengan model Maximin, Do Nothing 3 Dengan model Criterion of Realism, buat Large Plant 4 Dengan model Equally Like, buat Small Plant 5 Dengan model Minimax Regret, buat Small Plant

Hurwicz 0.8 124,000 76,000 124,000

h Tidak Menyesal)

Bayes Theorem for Thompson Lumber Example Probability Revisions Given a Positive Survey

State of Nature FM UM

P (Survey Positive| State Posterior of Nature) Prior Prob Joint Prob Probability 0.7 0.5 0.35 0.78 0.2 0.5 0.1 0.22 P(Sur.pos)= 0.45

Probability Revisions Given a Negative Survey

State of Nature FM UM

P (Survey Positive| State Posterior of Nature) Prior Prob Joint Prob Probability 0.3 0.5 0.15 0.33 0.8 0.5 0.4 0.89 P(Sur.neg)= 0.55