Decision Trees

• Decision trees are mathematical models that help management make decisions, particularly in situations of uncertainty or where there are multiple choices.
• Decision trees are sensitive to the assumptions that are made, such as probabilities and financial results.
• The expected value of a decision can be calculated by subtracting the cost of the option from the total value.
• The expected value of a decision can be calculated by multiplying the probabilities by the financial results.
• Decision trees work by using a series of assumptions, estimates, probabilities from which likely outcomes are calculated.
• Decision trees use estimates and probabilities, therefore when evaluating the outputs from the decision tree, it's important to question whether these estimates and probabilities have been made carefully and cautiously.
• The four steps in building a decision tree are: identifying options, adding possible outcomes, calculating costs, and calculating expected values.
• In a decision tree, the decision to be made is represented by a square and outputting from the square are the options available to you, typically as lines.
• In a decision tree, circles represent the uncertain outcomes, the possible outcomes.
• In a decision tree, the expected value is calculated by multiplying the probability of the result by the result itself.
• Probability in a decision tree ranges between one, which is certainty, and zero, which is no chance.
• In a decision tree, the probabilities and outcomes of all of them when added together should equal one.
• The expected value of an outcome is calculated by multiplying the financial results by the probability.
• The net gain is calculated by adding up all the expected values of each outcome and subtracting their cost.
• Decision trees are a part of the scientific decision-making process
• The limitations of decision trees are:

  • Uses estimates of the probability (could be wrong)
  • The value depends on how accurate the estimates are
  • Only uses financial and quantifiable data