CHAPTER 6 THE COMPLEXITY OF REAL-WORLD PROBLEMS 163
Figure 6–13
Schematic Tree for Diagnostics
limitations—correcting one error, waiting a long time for the tree to reevaluate,
and then finding another error.
After examining the sensitivity analysis results in Figure 6–12, the team
drew the schematic decision tree shown in Figure 6–13. These seven uncertain
variables account for almost all of the uncertainty in the three alternatives. All
other variables were set to their base-case values or determined by model
calculations. The decision node occurs before (to the left of) all the uncertainty
nodes because information on the resolution of these uncertainties will not be
available until after the decision is made.
The team started to more carefully examine the uncertainty on the
variables in the tree. The previous ranges were fine for sensitivity analysis, but
now the facilitator wanted to assess complete distributions to obtain better
quality information for the tree. (See Chapter 12 for more on this process.) Aftersome careful thought, however, the team judged that the information prepared for the sensitivity analysis represented their best state of information. They assigned .25/.50/.25 probabilities to the values used in the 10/50/90 rangesin the sensitivity analysis—see Chapter 2 for a justification of this process.
The one probability they were missing was the probability of technical
success for DiagStatic New development. A lengthy discussion of technical
hurdles with the design staff established that the probability of technical