Durgical
EN
Open the app

Probability rules: addition, multiplication, conditional and «at least one»

  • Concept

Genetics questions, screening results and repeated tests all come down to four small rules. Get them right and many exam calculations take seconds.

Updated

What it is

Probability runs from 0 (never) to 1 (always). The addition rule gives the chance of A or B; the multiplication rule gives the chance of A and B; conditional probability P(A | B) is the chance of A among those who already have B; the complement rule (1 − P) turns «at least one» questions into «none». Predictive values and post-test probability are conditional probabilities, which is why these rules sit under every diagnostic test.

Key facts

  • Independent means knowing B does not change the chance of A: P(A | B) = P(A).
  • Mutually exclusive means they cannot happen together: P(A and B) = 0 (blood group A or O in one person).
  • Complement: P(not A) = 1 − P(A); «at least one» = 1 − P(none).
  • Sensitivity is P(test + | disease); positive predictive value is P(disease | test +): same events, reversed condition.
  • Bayes' rule links them: P(A | B) = P(A) × P(B | A) / P(B), the engine of pre-test to post-test probability.

Worked example

  1. Made-up ward of 200 patients: 20 carry MRSA (0.10), 60 have diabetes (0.30) and 10 have both (0.05).
  2. Addition, overlapping events: P(MRSA or diabetes) = 0.10 + 0.30 − 0.05 = 0.35, that is 70 patients; the 10 with both are counted once.
  3. Conditional: P(MRSA | diabetes) = 0.05 / 0.30 ≈ 0.17 (10 of 60), above the ward-wide 0.10.
  4. Independence check: if unrelated, both would occur in 0.10 × 0.30 = 0.03 of patients (6 people); 10 are seen, so they are linked.
  5. General multiplication: P(both) = P(diabetes) × P(MRSA | diabetes) = 0.30 × (10/60) = 0.05, which matches the 10 patients.
  6. Mutually exclusive: if 0.40 of patients are group A and 0.45 group O, P(A or O) = 0.40 + 0.45 = 0.85, with no overlap to subtract.
  7. Complement: a healthy person has 10 independent check-up tests, each with a 5 % false-positive chance; P(all normal) = 0.95¹⁰ ≈ 0.60.
  8. P(at least one abnormal) = 1 − 0.60 = 0.40, not 10 × 0.05 = 0.50.
  9. Reading: subtract the overlap for «or», use the conditional for «and» when events are linked, and go through «none» for «at least one».

Look-alike ideas that exam questions swap

IdeaMeansFormulaClassic example
Mutually exclusiveCannot happen togetherP(A and B) = 0; P(A or B) = P(A) + P(B)One person's ABO group: A or O
IndependentOne does not change the other's chanceP(A and B) = P(A) × P(B)Each child of two carrier parents: 1/4 affected, whatever the siblings
Dependent (conditional)Knowing B changes the chance of AP(A and B) = P(B) × P(A | B)MRSA is commoner among the diabetic patients on this ward
P(B | A) vs P(A | B)Same pair, reversed conditionLinked by Bayes' rule through P(A) and P(B)Sensitivity P(+ | D) vs PPV P(D | +)

Markers

  • P(A or B)
  • P(A and B)
  • P(A | B)
  • 1 − (1 − p)ⁿ
  • Bayes

Common traps

  • Mutually exclusive events are independent: the opposite; if one occurs the other cannot, so P(A and B) = 0, not P(A) × P(B)
  • The chance of at least one false positive in 20 tests at 5 % is 20 × 5 % = 100 %: it is 1 − 0.95²⁰ ≈ 64 %; adding probabilities works only for exclusive events
  • P(disease | positive) equals sensitivity: that is P(positive | disease); the chance a positive patient is ill also depends on prevalence (Bayes)
  • P(A or B) = P(A) + P(B) for any two events: for overlapping events this counts the overlap twice; subtract P(A and B)

Test yourself

Think of the answer, then tap the question to see it.

Two parents are both carriers of an autosomal recessive disease. What is the chance that at least one of their 3 children is affected?

Each child has 1/4, independently. P(none affected) = (3/4)³ = 27/64, so P(at least one) = 37/64 ≈ 0.58.

In a town 20 % of adults smoke and 30 % have hypertension, independently. What share has both, and what share has at least one?

Both: 0.20 × 0.30 = 0.06. Either: 0.20 + 0.30 − 0.06 = 0.44; the overlap is subtracted once.

Why can two mutually exclusive events (each with a probability above 0) never be independent?

Independence needs P(A and B) = P(A) × P(B), a positive number; exclusive events have P(A and B) = 0. Knowing one happened tells you the other did not.

A test has 90 % sensitivity. Why is it wrong to tell a patient with a positive result that they are 90 % likely to have the disease?

90 % is P(positive | disease). The patient's question is P(disease | positive), the predictive value, which also depends on how common the disease is in people like them.

Add to my cards (4)

Connections

← All of epidemiology

Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.

Sources: 8.1.10. How can Bayesian methodology be used for reliability evaluation? (NIST/SEMATECH e-Handbook of Statistical Methods, National Institute of Standards and Technology, 2012); 8.1.8.1. Competing risk model (NIST/SEMATECH e-Handbook of Statistical Methods, National Institute of Standards and Technology, 2012); Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests - Guidance for Industry and FDA Staff (U.S. Food and Drug Administration, 2007).

Spotted a mistake?We are always improving for the community. Tell us what is wrong and our team will review it.Report a mistake

Representative model, for educational purposes only.