# Decision Theory and Phase

Topics: Decision theory, Decision tree, Decision making Pages: 7 (2241 words) Published: January 10, 2013
Executive Summary (Farzan's Part)

‍Introduction To Company and Problem (Farzan)

‍Methodology

The decision analysis that we did encompassed a number of different aspects. First we had to describe what the optimal decision would be based on the probabilities and the payoffs. For this, we designed a decision tree model to reflect the decisions that we to be made at each step, furthermore, the events that are out of control of the company are represented by chance nodes. After completing the outline of the model, we proceeded to “foldback” the tree to find out which decision would give us the highest Expected Monetary Value (EMV) calculated using the following formula: EMV = PAMA + PBMB (Dont worry about the formatting yet...we will do that section later)

Where:
PA = Probability of event A
PB = Probability of event B
MA = Payoff of event A
MB = Payoff of event B

Based on this analysis, we have can make a decision that will maximize our long term expected monetary value. Further, the decision tree allows the user to visualize the decision outcomes and the associated likely hoods involved with each decision.

The concept of risk was an important conclusion of the decision tree. After discussion, we contacted the consulting firm of Foresight Consulting to help manage this risk. They informed us that they have a method to determine the outcome of Phase I and Phase II before actually conducting these expriements, however, they did not quote a price. To further advance this analysis, we calculated the Expected Value of Perfect Information (EVPI). This figure tells Merck the maximum they should be willing for Foresight's services.

Another method of examination was to consider a sensitivity analysis. We used the existing decision tree model and applied a Monte Carlo simulation on some of the variables. The Monte Carlo simulation is used in stochastic and nondeterministic fields by using a pseudorandom algorithm to calculate a forecast cell. A total of three simulations were run. In the first simulation we varied costs associated of each Phase and the payoffs of successful drug release. For the second simulation we varied the probabilities accompanying each successful outcome. And the third simulation was a combination of the last two where we varied both the dollar amounts and the probabilities associated with each outcome. For each of the simulations we were given a mean and standard deviation for the forecast cell. Furthermore, we were able to perform a sensitivity analysis that shows which variable affects the forecast cell the most. Based on these sensitivity numbers we came up with ways so that Merck can control their risks and avoid losing too much money.

‍Analysis and Key Findings

(Need a transition paragraph)

‍Decision Tree
Expected Monetary Value
Two possible alternatives were examined: to buy the patent for KL-798 or not to buy. If the drug is purchased and Merck follows through with the research then the Expected Monetary Value of this option would be a loss of \$ 260 000.00. This Criterion is calculated using the decision tree in Appendix A Decision Tree Diagram on page ZZ (When Referencing to the appendix we need to make sure we mention the page number).

The Decision tree shows what decision should be made given the circumstances. First, assuming that Merck will proceed with the purchase of KL-798, Merck will have to make a choice as to whether to complete Phase I, which only has 60% chance of success and will cost 5Million dollars. If they pass phase I, according to the EMV criterion, it would be advisable to continue on to phase II which has a number of different outcomes. Firstly, there is a chance that they could cure obesity only, cholesterol only, or both cholesterol and obesity.

If phase II gives an indication that it cures obesity, then Merck should continue on with Phase III and seek FDA approval. If phase II gives an indication of cholesterol...