Exercises 10.2 to 10.11 are based on past Oxford University examination papers reprinted in Applied Decision Analysis by Derek Bunn (McGraw-Hill, 1984)

10.2 A company has to decide whether or not to drill for oil in a particular place. There are three possible results of drilling: a high yield worth \$10 nillion, a moderate yield worth \$5 million, or no oil. The drilling operation costs \$5 million. At similar places, 50%, 30%, and 20% of previous drillings have given high. moderate. or no yield. respectively. A seismic test is available which would indicate a favorable. neutral. or discouraging prospect for the drilling.

There is a 52% chance the test will give favorable results;
if it does, the respective ;probabilities of High, Moderate, or No yield become 35/52, 15.42, and 2/53

There is a 22% chance the test will give favorable results;
if it does, the respective ;probabilities of High, Moderate, or No yield become 10/22, 6/22, and 6/22

There is a 26% chance the test will give favorable results;
if it does, the respective ;probabilities of High, Moderate, or No yield become 5/26, 9/26, and 12/26

A. Find the best decision rule on the assumption that a test is made, and the maximum amount that it would be worth paying for such a test. If this is the actual cost of the test, what is the expected value of the potential well?

B. Past experience indicates the following probabilities for the results of the seismic test for the three different ultimate possible yields:

 Yield High Moderate none Favorable 0.7 0.5 0.1 Results of test Neutral 0.2 0.2 0.3 Discouraging 0.1 0.3 0.6

Show how  the probabilities that each test result will occur can be deriver from the priginal probabilities of each yield level.

C. Use your work for part B to show hoe the probabilities after each possible test result is observed can be calculated.

10.3 A company is deciding whether to launch or abandon a new product, The future profits if the product is a success amount to an estimated £10 million, compared with a loss of £5 million if it is a disaster. Abandoning the project at this moment in time has no implications for future cash flows, since all the costs have already been sunk in the project. The decision maker estimates that there is a 0.3 chance of the project being a success compared with 0.7 for a disaster. Should the decision maker launch or abandon the new product? What is the most he or she should pay for a perfect predictor?

Instead of launching the product nationally, a regional test market of the product could be undertaken. This would cost £1 million and would give some indication of the national market for the product, If the product were a success in the region, then £1.2 million would be recovered on the test, compared with nothing if it were a regional disaster Past experience with this regional test market is such that there were regional successes 90 percent of those times in which the products were national successes and regional failures 80 percent of those times in which the national launches were failures. Should the decision maker go ahead with the test market ?

10.4 A satirical magazine specializes in publishing embarrassing details of the private lives of eminent citizens. The editors are aware of a scandalous rumor about a prominent politician and wish to proceed with the story in a manner which maximizes expected net income The risk of an expensive libel action cannot be ignored and the editors estimate that the effects of such an action will be a cost of £125,000 net if lost, taking into account possible fines, legal costs, effects on future sales, and so on, and zero if won. These costs, and the probability of losing, estimated to be 0.2, are, they believe, independent of whether or not the rumor is true. However, the probability of the politician suing does depend on whether the rumor is true a nd they estimate that the probability of his taking legal action is 0.3 if the rumor is true and 0.9 if it is not. They currently estimate the probability of the rumor being true at 0 5.

They have three possible courses of action: (1) dropping the story, (2) publishing immediately or (3) hiring a private detective to investigate the rumor further so that after his report they could take a decision on whether or not to publish. The private detective they have in mind would cost £10,000, and he is good but not infallible. The probability of his saying the rumor is true is 1.0 if indeed it is true and 0.2 if it is false. He will definitely answer in one direction or the other, but t he probability of being sued depends only on whether the rumor is true or not.

They estimate extra net income (gross of any costs of legal activity) from increased circulation if they publish to be £50,000, and this would be independent of the truth of the story. What should they do? How much should they be prepared to pay for a report from an infallible private detective ?

10.5 Manufacturers of a range of plastic components, pipes, guttering, etc., for the building industry estimate that they face a 20 percent drop in sales within three years unless they can expand or improve their current product line. Accordingly, the research manager has been given £12,000 extra in the annual budget and has ascertained that sales of piping can be greatly increased if a specified improvement in the strength and durability of current products can be achieved at reasonable cost. Re search on piping is currently being undertaken by two teams whose approaches are entirely independent, with a view to achieving a definite result by the end of the current year. The manager estimates the chances of success within the year as 50 percent for a group of chemists experimenting with a low-cost material which would improve the final product. His corresponding estimate for a group of engineers attempting to simplify an excessively costly production process by which piping of the required quality is already made is 60 percent.

With the extra funds the manager could expand the research teams. He can hire chemists at an additional cost (including non-salary expenses) of £4000 per year and hire engineers for a similar cost of £6000 per year. He estimates that another chemist would reduce the chances of that team's failure to 40 percent, two extra would reduce it to 35 percent, and three extra to 30 percent. Another member of the engineering team would reduce that team's chance of failure to 30 percent, while two extra would reduce it to 25 percent.

Alternatively, he can second consultants from an associate company at an extra cost of £3000 per year each to explore fiber reinforcements of plastic pipes. A minimum-size team of two persons is estimated to have a 50 percent chance of success within the year, but this would be increased to 60 percent and 70 percent by a three or four person team.

Which new researchers should the manager hire, and what is the resulting chance o£ success ? What is the expected value o£ perfect information ?

10.6 The government has spent £1,000 million on developing a new aircraft which is suitable for only one route. Unfortunately, the aircraft is rather noisy and is in danger o£ being refused landing rights on this route. There are two possible courses o£ action. The government can apply for rights now. Or it can commission further investment in an attempt to make the engines quieter before an application is submitted. There is no possibility of applying for rights more than once since if they are refused, the political situation will be such that the decision will not be reversed.

The research and development to make the engines quieter would cost £100 million, and the current estimate o£ the probability o£ success is 0.7.

Before embarking on this further project, the government would carry out a pilot project for £10 million. The report of the pilot project would indicate that, for £100 million, either the quiet engines are certain or they are unlikely. The engineers on the project are biased toward aircraft development, so the government estimates that even i£ the large project will fail to produce quieter engines, there is still a probability o£ 0.3 of the pilot project report saying that they are certain. 1f the larger project will in fact succeed, the pilot project is certain to indicate this. If the pilot project indicates that success is unlikely, then the government will not proceed with the main project and will apply for landing rights with the existing engines. If the pilot project report is favorable, the government will proceed with the larger project.

If quiet engines are successfully developed, landing rights are certain. If they are not developed, there is only a 50:50 chance o£ receiving rights. With either quiet or noisy engines the present value of operating profits on the route is £3 00 million.

What should the government do? What is the most it should pay to buy off the groups opposing granting of the rights so that the application with the noisy engines would be certain? What should the government pay (assuming that the opposition is not bought off) above the £10 million for a pilot project which would predict with complete accuracy what the outcome of the main project would be ?

10.7 The marketing manager of a firm is considering the marketing of a newly developed product. He feels the current odds on its success are 4: 1, and he has already commissioned a market survey at a cost of £90,000, which is currently in progress . If the product is ultimately to be a failure on the market, it is estimated that the survey will indicate this with probability 4/5. However, there is an estimated 1/10 chance of a spurious indication of non-marketability. When the results of the first survey are available, he is considering a second smaller survey at a cost of £45,000, with a different statistical design, which has an estimated 2/3 chance of correctly predicting non-marketability when the first survey did, but an estimated 9/10 chance when it did not. On the other hand, it has an estimated 1/9 chance of spuriously indicating non-marketability when the first survey did, but a 1/12 chance when it did not. The company has invested £100,000 in the development of the product, and it is estimated that production and marketing cost s will lead to a loss of £500,000 if the product is a failure. On the other hand, its success is thought to be worth £1 million in profit to the company. What policy should the marketing manager recommend to the directors on current information ? Discuss briefly the usefulness of utilities in such situations.

10.8 The survey department of a ministry announces that a certain area of government land has property X. If the land does have property X, the probability of finding oil in marketable quantities is 9/10, but if it doe s not, the probability is only 1/10. This particular department has been wrong before, however, so the minister puts a probability of only 3/4 on the survey department's announcement being correct.

The government has an immediate choice between three actions. It can sell the land at £300,000. It can drill itself at a cost of £400,000. It can commission a specialist constant do a test on the land at a cost of £20,000. The test can have two outcomes, positive or negative, and is such that the probability of a positive outcome is 9/10 if X obtains and 3/10 if X does not obtain. After the test the government will be able to drill or sell the land. The land can also be sold if drilling fails.

The selling price of the land if the government drills and fails is £100,000. The government knows that it cannot conceal the result of the specialist consultant's test and predicts that the test is negative, the selling price will be only £1 00,000, but if the test is positive, the price will be £440,000.

The minister believes that the successful finding of oil is worth £1 million. What is the best policy if the government wishes to maximize expected monetary value ? What is the most the government should pay for an envelope whose contents would re veal, to itself alone, whether or not the land had property X?

10.9 An electronic system consists of two principal components. Failure of either can cause failure of the system. An a priori estimate of their individual reliabilities is 0.5 and 0.6, respectively, as they come of the production line, but in assembly there is a 50 percent chance of causing a failure to either, independent of their individual reliabilities. Either component may be tested separately before assembly, at respective costs of £50 and £60, but there is then only a 0.6 and 0.8 chance, respectively, of finding a fault leading to system failure during a specified period of use. The assembled system can be tested at a cost of £120, and in this event any faults are detected with certainty. The system is installed in a piece of oceanographic equipment, and its subsequent failure during the period of use is estimated to cost £200 in data loss, inspection, and repair.

Use decision analysis to determine the optimal pre-installation testing procedure. How would a re-estimate of failure cost at £300 modify your solution ? (Note: "Reliability " is probability of working.)

10.10 A manufacturer is planning a production facility which involves the machining of rough castings. Each casting costs him £5. He intends to use a process which costs £10 to machine each casting and gives a probability of 0.9 that satisfactory results will be obtained from a perfect casting, but some of the castings contain flaws, and the probability of success from a flawed casting is only 0.2. A second, more expensive, process is also available for which the probabilities of producing satisfactory results are 0.95 and 0.3 for perfect and flawed castings, respectively.

(a) If 2% of the castings are flawed, how high could the cost per casting of the second process be

(b) Suppose that the unit cost of process 2 is £0.85 greater than that of process 1, and each casting rejected before machining costs £5. In order to try to reduce the cost of scrap, the castings may be inspected for flaws before machining. This process would cost £2.50 per unit inspected and is not perfect:

the probability of accepting a flawed casting is 0. 10

the probability of rejecting a perfect casting is 0.02

There are four options to consider -- process 1 and process 2, each with and without preliminary inspection. For what values (if any) of P (the probability of a casting containing a flaw) is each option the most appropriate? You may wish to draw a graph.

10.11 A company has an embryonic idea for a new type of aeroplane engine. The successful development of the engine is much more likely if a recently developed substance has a certain property (call it property A). The managing director currently believes t hat the probability that this new substance has property A is ½. He approaches the government for permission to go ahead with a project for developing the new engine. The government replies that permission will only be granted if the company's research department produces a positive test for property A. It costs £40 million to carry out this test. If the substance does have property A, the test will certainly be positive, but if it does not have property A, there is still a 1 in 5 chance that t he test will be positive.

If the test result is positive, the managing director can send the idea to his main development department to turn it into an engine. The cost of this development program is estimated at £244 million. The test carried out by the research department is such that if the substance does not have property A but the test was positive, then the probability of successful development in the main department is 1/10. On the other hand if property A does hold, the probability of successful development in this department is 9/10. If development fails in the main department the manager can send the idea for development to an expert problem-solving unit at an estimated cost of £180 million. If the main department failed to solve the problem, the expert unit has probability ½ of solving it if A holds, but only 1/20 if not.

If the estimated gain (before taking into consideration the costs described above) from successful development is £840 million and the firm wishes to maximize the expected money value (EMV) find its optimal strategy (assuming that it can give up at any stage without extra cost) and its expected money value.