QR 26: CHOICE AND CHANCE
Assignment # 3: Trees Solving and Sensitivity Analysis
- Study Chapters 4 and 5 in Making Hard Decisions (MHD). While it is good to know how to make and read "tornado diagrams," you do not have to pay them much attention beyond that since we will take a different approach to sensitivity analysis later. Compare MHD with the excerpts in our Sourcebook from Sections 3.3 to 3.5 of Clinical Decision Analysis by Weinstein and Fineberg.
- Use the PrecisionTree or TreePlan Add-Ons for Excel to draw and solve decision trees. Precision Tree software and tutorials are accessible from Windows machines in the Science Center computer laboratories. In case that program does not load from within the "Courses" Folder, you can also open "Windows NT Explorer" (found in the Start Menu under "Programs") and go to the S: drive (also shown as "bin$ on 'Fas_nt1'"). In the S: drive, open the "Dtools" folder, then the "Ptree32" folder. Double-clicking on "ptree.xla" in the "Ptree32" folder will open Precision Tools. Check out what the tutorial says about the Texaco case, for example, and compare it to our treatment in class. See one of the course staff this week, especially if you need computer help.
- Problem Set from MHD to hand in:
- #5.7, 5.8, 5.9, 5.10
- Read the "Strenlar Case" (pp. 167-169). You do not have to work it all through, but comment on what kinds of data or assumptions you would need beyond what you are told in order to analyze this decision problem.
- Prepare notes that you can discuss and hand in concerning:
- "GPC' New Product Case" (pp. 163-165). You do not have to do the influence diagram (question 1) unless you find it helpful.
- "Job Offers Case, Part II" (pp. 214-215). Try to analyze this also using the "Method of Equal Swaps" you read about in Smart Choices. Extra Credit: Look up and critique how magazines actually do rate the desirability of different cities as places to live.
- Think about (but you don't have to hand in) questions 5.1 and 5.5 concerning the role of sensitivity analysis in decision theory. What are the strengths and weaknesses of tornado diagrams, for example? Sensitivity analysis usually concerns relatively small changes in EMV. Do your reservations about EMV as a decision criterion bother you more or less in that case? (Hint: What is differential calculus about?)