Solving the Decision Tree
Decision analysts are divided as to whether solving the decision tree or running a Monte Carlo simulation is the better way to develop an s-curve. Both methods work. If you have time, using both methods has the advantage of providing a step of quality assurance versus an input or computational error. We’ll start with solving the decision tree.
To solve the tree, we
• build out the tree using each uncertainty with corresponding p10/50/90 branches,
• compute the measure of value (usually NPV) and probability for each end point branch,
• build a table of these results,
• sort from high to low NPV,
• calculate the cumulative probability of each branch, and
• plot the cumulative probability versus the sorted values using ...