A6: Metrics for Decision Support: Causal Models

DUE: 11:59PM, Friday 20 October 2017

20 points

1. Problems

  1. (10 points) It is known that a particular type of software test is certain to identify virus X if it has been inserted into a computer system. However, there is also a 5% probability of a false alarm (i.e. there is a 5% probability that the test will be positive when virus X has not been inserted). You run the test and the outcome is positive.
    1. What can you conclude about whether or not the system really is infected with the virus X?
    2. How would your answer to (a) change if, additionally, it was known that virus X had been inserted into approximately one in every thousand computer systems?
    3. Suppose the only known effective fix for virus X costs $250,000 when the full costs of system shutdown and repair are considered. From a risk assessment perspective what action would you recommend? You should state any assumptions about additional information needed.

  2. (10 points) Look again at the story recounted by the keynote speaker on the first page of chapter 7 of the Fenton and Bieman text. Draw a Bayesian Network (BN) model (with 5 nodes) that "explains" the phenomenon he observed. See Figure 7.8 in the text for an example BN model. (You do not need to include probabilities in your model.)

2. Submission

  • Submit your answers in one PDF file called a6.pdf.
  • Put your name at the top of the each page.
  • Important Reminder: You may use material from the book or other sources in your answers. However, you must cite your sources properly. Any verbatim quotations must be enclosed in quotation marks, with page numbers indicated. You will receive severe point deductions if you use material from the text or other sources that is not properly cited.