University of Sydney · FACULTY OF STATISTICS

QBUS5001 Chap.3 Probability Rules and Bayes Reasoning

- one subject, every graph, every model, every mark
6 Chapters5-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 3 of 10 · QBUS5001

Probability Rules and Bayes Reasoning

This chapter translates business wording into events before applying probability rules. It distinguishes joint, marginal and conditional probability, uses complements and independence carefully, and builds Bayes updates from all routes to observed evidence. The decision layer explains why sensitivity is not posterior probability and why a low base rate can make false positives dominate a flagged group.

A joint probability sits in the overlap of events, a marginal probability totals across the other classification, and a conditional probability changes the reference group to cases satisfying the condition. The addition rule subtracts an overlap counted twice, while the multiplication rule builds a joint event from a starting probability and a conditional branch.

Independence is a testable relationship between probabilities, not a casual synonym for unrelated. Bayes reasoning reverses a conditioning direction by combining prior prevalence with evidence rates. The denominator must include every mutually exclusive route to the observed evidence.

In screening and risk settings, the false-positive route can be numerically larger than the true-positive route because it operates on a much larger base population. The posterior result becomes decision-ready only after it is compared with the cost or threshold attached to action.

In this chapter

What this chapter covers

  • 01

    Random experiments, outcomes, sample spaces and events

  • 02

    Complements, unions and intersections

  • 03

    Joint, marginal and conditional probability

  • 04

    Independence versus mutual exclusivity

  • 05

    Total probability and Bayes theorem

  • 06

    Base rates, screening accuracy and decision thresholds

Worked example · free

Find defect probability after a flag

Q [5 marks]. Five percent of parts are defective. A screen flags 88% of defective parts and 6% of acceptable parts. What is the probability a flagged part is defective? The mark allocation used here is not a University assessment scheme.
A probability tree separates prior rates and conditional ratesBranches show a low base rate, a high sensitivity and a false-positive path leading to flagged cases. Claim Error 6%prior rate No error 94%complement Flagged 90%true positive Flagged 8%false positive
The two flagged paths must be weighted by their prior branch sizes, so a high sensitivity does not by itself equal the probability that a flagged claim contains an error.
  • +1Compute the defective and flagged path: 0.05 times 0.88 equals 0.044.
  • +1Compute the acceptable and flagged path: 0.95 times 0.06 equals 0.057.
  • +1Add both routes to obtain P(flagged)=0.101.
  • +1Divide 0.044 by 0.101 to obtain approximately 0.4356.
  • +1Interpret the posterior and distinguish it from 88% sensitivity.
About 43.56% of flagged parts are defective because false positives act on the much larger acceptable group.
Sia tip — The denominator in Bayes theorem contains every mutually exclusive route to the evidence you observed.
Glossary

Key terms

Sample Space
The set of all possible outcomes of a random experiment at the level used in the analysis.
Joint Probability
The probability that two or more stated events occur together.
Conditional Probability
The probability of an event within the restricted sample space created by another event.
Bayes Theorem
A rule for updating a prior probability after evidence by weighting all routes to that evidence.
False Positive
A positive signal produced for a case that does not have the target condition.
FAQ

Probability Rules and Bayes Reasoning FAQ

How do I recognise conditional probability?

Look for wording such as given, among, of those or if. The denominator becomes the conditioning group rather than the whole sample. Define the conditioning event in words before calculating so the final sentence does not reverse the probability.

What is the difference between independence and mutual exclusivity?

Independent events do not change each other's probability and may occur together. Mutually exclusive events cannot occur together. Except for zero-probability cases, mutually exclusive events are therefore not independent because observing one rules out the other.

Why can a high-sensitivity screen have a low posterior probability?

Sensitivity applies among cases that already have the condition. Posterior probability applies among flagged cases. When the condition is rare, a modest false-positive rate acts on a much larger group and can produce more false flags than true flags.

Study strategy

Exam move

Translate every question into named events and draw a two-way table or tree before using a formula. Practise saying both directions of a conditional probability aloud. For Bayes questions, compute each joint path, add every route to the observed evidence and interpret the posterior as a frequency among flagged cases. Finish with the business threshold or cost that turns probability into a decision.

Train conditional direction with matched pairs: write P(flagged given defective) beside P(defective given flagged), then describe the different denominator in ordinary words. Convert one tree into a two-way table and verify that both representations give the same marginal and posterior probabilities. When a complement appears, identify the complete reference group before subtracting from one.

For independence, calculate the product of marginals and compare it with the joint probability instead of relying on the story alone. Vary only the base rate in a screening example and watch the posterior move while sensitivity and specificity stay fixed.

That sensitivity analysis makes the base-rate effect memorable and prepares you to explain why a seemingly accurate screen may still produce many false alarms in a low-prevalence population. Before finishing, test the result with natural frequencies. Apply the prior rate to a convenient population, split it into target and non-target groups, then apply the conditional evidence rates.

The posterior fraction among all evidence-positive cases should match the formula. If it does not, the error usually lies in a missing branch or a reversed conditioning direction.

Working through Probability Rules and Bayes Reasoning in QBUS5001? Sia is AskSia’s AI Statistics tutor — ask any QBUS5001 Probability Rules and Bayes Reasoning question and get a clear, step-by-step explanation grounded in how QBUS5001 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 202 of your subjects at University of Sydney, plus 1,000+ more Bibles across the AskSia library.
Sia - your QBUS5001 tutor, unlimited, worked the way the exam marks it
The full 5-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works
QBUS5001 · Foundation in Data Analytics for Business - independent study guide on the AskSia Library. More University of Sydney subjects · Microeconomics across all universities