Lesson 8 of 8 · 16 min
Modelling and solving probability problems
Choose a representation, state your assumptions, solve unfamiliar probability problems with and without technology, and check that your answers are reasonable.
What you will be able to do
- Choose a Venn diagram, two-way table, tree diagram or rule to suit a problem.
- State the assumptions in a probability model and evaluate them.
- Solve multi-step problems with and without technology, and check that answers are reasonable.
Real problems don't arrive labelled "use a tree diagram". A medical screening test, a State of Origin series, the smoke alarms in a house: in each one you decide how to represent the situation, what to assume, and whether your answer makes sense. This lesson puts the whole topic to work on problems like these.
| If the information looks like… | Try |
|---|---|
| two or three overlapping groups, with counts or probabilities for "and", "or", "neither" | a Venn diagram |
| two events and their complements, often survey data | a two-way table |
| stages in order, or "if … then the probability is …" | a tree diagram |
| probabilities of events given directly, with one unknown | the complement, addition or multiplication rule |
| "at least one" of several independent events | the complement: |
Solving a probability problem
Formulate
Name the events with letters, write what you are given in probability notation, choose a representation and state any assumptions, such as independence.
Solve
Apply the rules, by hand or with technology. Keep exact values until the end.
Evaluate and verify
Are all probabilities between 0 and 1? Do branches and tables add up? Does the answer fit your intuition, or a quick simulation?
Communicate
Answer the question asked, in context, rounded as required, and say what your assumptions mean for the answer.
Here is a classic problem that most people, including many doctors, get wrong. A screening test is accurate in the sense that it is positive for 95% of people with a condition. So if you test positive, is it 95% likely that you have the condition? Not even close.
Tick the two paths ending in Positive to find . How much of that probability comes from people who have the condition?
Worked example
How reliable is a positive test?
A screening test for a medical condition is used in a population where 2% of people have the condition. If a person has the condition, the probability that the test is positive is 0.95. If a person does not have the condition, the probability that the test is positive is 0.08.
(a) Find the probability that a randomly chosen person tests positive.
(b) A person tests positive. Find the probability that they have the condition, correct to three decimal places.
(c) Comment on what this means for a person who receives a positive result.
Step 1: Name the events and write what you know
Let be "has the condition" and be "tests positive". Then , and .
Step 2: Draw the tree
First stage: 0.02, 0.98. After : 0.95, 0.05. After : 0.08, 0.92.
The given test probabilities are conditional on whether the person has the condition, so condition status goes first.
Step 3: Multiply along the two positive paths
and .
Step 4: Add them for (a)
Step 5: Reverse the condition for (b)
Step 6: Check with a population of 10,000
About 200 people have the condition and 190 of them test positive. About 9800 don't, and 784 of them test positive. Of the positives, have the condition.
Imagining a large round population is a quick, reliable check on a conditional probability.
Step 7: Interpret for (c)
Only about 1 in 5 positive results comes from a person with the condition. Because the condition is rare, most positives are false positives from the large group without it, so a positive screening result should be followed by a more accurate diagnostic test.
Try it yourself first, then check the working one step at a time.
Answer
(a) (b) (c) About 80% of positive results are false positives, so a positive result means "test again", not a diagnosis.
Method
Define events and identify which given probabilities are conditional.
Tree: condition first, then test result.
Add the paths for , then reverse with .
Interpret in context and check with a round-number population.
Watch out for
Answering 0.95, which is , not .
Forgetting the false-positive path when finding .
The same test is used in a group where 10% of people have the condition. As before, and . Find the probability that a person who tests positive has the condition, correct to three decimal places.
Independence is the assumption behind every "backup" system. If each of several smoke alarms fails independently, the chance that all of them fail is a product of small numbers, which becomes tiny quickly. The complement rule then gives the chance that at least one works.
Worked example
How many smoke alarms?
Each smoke alarm in a house works when needed with probability 0.85, independently of the other alarms.
(a) Find the probability that at least one of two alarms works.
(b) Find the smallest number of alarms needed so that the probability that at least one works is at least 0.999.
(c) Comment on the assumption that the alarms work independently.
Step 1: Use the complement for (a)
Each alarm fails with probability . By independence, both fail with probability , so .
Step 2: Write the condition for (b)
With alarms, . We need , that is, .
Step 3: Test values of
, which is more than 0.001. , which is less than 0.001.
Powers of 0.15 get smaller as grows, so the first that works is the smallest. A calculator table or spreadsheet does this quickly.
Step 4: Conclude
4 alarms are needed: , but 3 alarms give only .
Step 5: Evaluate the assumption for (c)
Alarms in the same house might fail together: they could share a power circuit, have flat batteries if nobody tests them, or come from the same faulty batch. If failures are linked, the chance that all of them fail is higher than the model says, so the real reliability is lower than 0.9995.
Try it yourself first, then check the working one step at a time.
Answer
(a) 0.9775 (b) 4 alarms (c) Independence may not hold, because alarms can share causes of failure; then the model overstates the reliability.
Method
For independent failures, is a product of the failure probabilities.
For "smallest ", test values (by hand or with a table) until the condition is first met.
Comment on whether independence is realistic.
Watch out for
Using , the probability that all the alarms work, instead of the complement of all failing.
Rounding up to 0.999 and choosing 3 alarms.
Without technology
Work out powers by hand until the condition is met:
Keep the exact decimals, and write a sentence that compares each value with 0.001. On Paper 1, numbers are chosen so this stays manageable.
With technology
Enter in a graphics calculator table, or fill a spreadsheet column, and read off the first whole number that gives at least 0.999. A simulation, using random numbers to run thousands of imaginary trials, is a quick check on a tree-diagram answer: if about 19.5% of simulated people who test positive have the condition, your calculation is probably right.
Four friends each enter a different competition. Each has a probability of 0.27 of winning, independently of the others. A student calculates the probability that at least one of them wins as . What has gone wrong?
A factory has two machines. Machine A makes 60% of the items, and 3% of its items are faulty. Machine B makes the rest, and 5% of its items are faulty. Which representation shows the given information most directly?
Common mistake
Ignoring the base rate
and are different. When a condition is rare, most positive results come from the large group without the condition, so can be small even for an accurate test.
Exam tip
Complex unfamiliar questions
Define your events, state the assumptions you make (such as independence), show the model you build, and finish with a sentence in context. When asked to evaluate, comment on whether the answer is reasonable and on the limitations of your assumptions, for example "if the games are not independent, the true probability may be higher or lower".
Summary
Choose the representation from the shape of the information: groups → Venn, survey → table, stages or "if … then" → tree.
State assumptions such as independence or equally likely outcomes, and judge whether they are realistic.
Reverse conditionals carefully: can be far smaller than when is rare.
For "at least one" of several independent events, use , never a sum of probabilities.
Check every answer is a probability between 0 and 1 that makes sense in context.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.