Statistics track

Statistics · Phase 2

Probability Foundations

Random variables, Bayes, key distributions, CLT and LLN.

In plain English

Probability is bookkeeping for uncertainty. It keeps your beliefs consistent so that the numbers you quote cannot contradict each other.

The advanced view

The axioms (non-negativity, normalisation, additivity over disjoint events) generate everything else. Conditional probability defines dependence, expectation is a linear operator regardless of dependence, and variance is not — which is exactly why covariance appears in portfolio mathematics.

Probability is the language of uncertainty, and finance is the business of pricing uncertainty. A random variable maps outcomes to numbers: discrete variables have probability mass functions, continuous ones have densities, and the CDF gives P(X ≤ x) for both. Bayes' theorem updates probabilities with new evidence — it is how credit models revise default probabilities, how fraud detection works, and how diagnostic tests are interpreted.

P(A|B) = P(B|A) × P(A) / P(B)
E(X) = Σ p_i × x_i
Var(X) = E[(X − μ)²] = E(X²) − μ²
P(A∪B) = P(A) + P(B) − P(A∩B)

Key distributions: normal (returns over short intervals), log-normal (prices, because returns are roughly normal and prices cannot go negative), binomial (successes in n trials, and the basis of binomial option pricing), Poisson (rare-event counts such as defaults or claims) and Student's t (heavier tails, for small samples and fat-tailed returns). The Central Limit Theorem says sample means converge to normal regardless of the underlying distribution; the Law of Large Numbers says they converge to the population mean. CLT tells you the shape, LLN the target.

Essential vocabulary

Expected value E(X)
Probability-weighted average of outcomes. The long-run average.
Variance Var(X)
Expected squared deviation from the mean. Standard deviation is its root, in the units of X.
Covariance & correlation
Covariance measures co-movement; correlation normalises it to [−1, 1]. Critical for portfolio construction.
Conditional probability
P(A|B): probability of A given B has occurred. Changes the sample space.
Independence
P(A∩B) = P(A)×P(B). Most financial assets are not independent.

Why it works

Bayes' theorem works because it is just the definition of conditional probability read both ways: P(A|B)P(B) = P(A∩B) = P(B|A)P(A). It corrects for base rates because the denominator counts every way the evidence could have arisen, not only the way you had in mind — the reason rare-condition test results are so often misread.

Common pitfalls

  • ×Ignoring the base rate when a test is accurate but the condition is rare.
  • ×Adding probabilities of events that are not mutually exclusive.
  • ×Assuming independence to make the arithmetic easy when the case implies correlation.

How it is used — a screening test in your head

Step 1 of 4

  1. 1Prevalence 1%, sensitivity 99%, false positive rate 5%.

Deeper

Deeper: Bayes, base rates and conditional thinking

P(A|B) = P(B|A) × P(A) ÷ P(B). The prior P(A) — the base rate — is what people drop, which is why a 99%-accurate test for a 1-in-1,000 condition still produces mostly false positives. In business the same trap appears in fraud screening, churn prediction and due-diligence red flags.

Expected value alone is not a decision rule when outcomes are lumpy: the variance and the possibility of ruin matter. That is why firms buy insurance with a negative expected value, and why 'positive EV' bets with a 5% chance of bankruptcy are declined.

Must know cold

  • P(A|B) = P(B|A)P(A) ÷ P(B).
  • Independent events multiply; mutually exclusive events add.
  • E[X] = Σ p·x; Var(X) = E[X²] − (E[X])².
  • Base rates dominate when the event is rare.

Exercises

Try each one on paper before revealing the worked solution.

Exercise 1

A fraud model flags 1% of clean transactions and catches 95% of fraud. Fraud is 0.2% of volume. A transaction is flagged — how likely is fraud?

Exercise 2

A project pays +100 with probability 0.6 and −80 with probability 0.4. Would you run it once? Fifty times?

Figure — where the 1.96 comes from
95% inside ±1.96σ, 2.5% in each tailstandardised outcomedensity

The critical values you quote are just cut-points on this density: ±1.645 for 90%, ±1.96 for 95%, ±2.58 for 99%. Knowing the shape means you can sanity-check any z-score or p-value in your head.

References

  • Newbold, P., Carlson, W. L. and Thorne, B. M. (2020). Statistics for Business and Economics. 9th Edition, Pearson, Harlow.

Statistics glossary for this phase

The terms an interviewer expects you to use precisely — with the pitfall attached to each.

Expected value

Probability-weighted average of the outcomes.

In finance

The number behind decision trees, expected loss and any scenario-weighted valuation.

Pitfall

×Treating the expected value as an outcome that will actually happen; nobody earns the average of a coin flip.

Conditional probability

Probability of A given that B happened.

In finance

Default given a downgrade, churn given a price rise, fraud given an alert.

Pitfall

×Swapping P(A|B) with P(B|A) — the classic base-rate error in screening and credit tests.

Normal distribution

Symmetric bell curve summarised by mean and standard deviation.

In finance

The default assumption behind volatility, VaR and option pricing.

Pitfall

×Financial returns have fat tails; the normal curve understates crash frequency badly.

Practise this

The drills and cases where this phase turns into arithmetic you do out loud.