Always compute the denominator (P(A)) as the sum of matches.
Chapter 3: Identify the Right Model
Binomial
Fixed trials n, constant success prob p.
Counts successes in n.
Example: number of sixes in 50 rolls.
Poisson
Counts events in time/space with rate λ.
Unbounded counts: 0,1,2,…
Example: calls in the next hour.
Geometric
Trials to first success; constant p.
Support: 1,2,3,… (includes the success trial).
Example: rolls until first “1”.
Uniform (continuous)
All values equally likely on [a,b].
Density: f(x) = 1 / (b − a).
Example: arrival time uniformly in a 10-min window.
Model ID strategy: Ask 3 questions — (1) Are outcomes discrete counts or a continuous value? (2) Is there a fixed number of trials n with success probability p (Binomial) or a rate in time/space λ (Poisson)? (3) Are you waiting for the first success (Geometric)?
Calculation Part — A/B/C
Part A: Bayes’ Theorem
Workflow
Set up rows for categories Bi with prior P(Bi) and likelihood P(A|Bi).
Compute matches: mi = P(A|Bi)P(Bi).
Sum matches: P(A) = Σ mi.
Posterior: P(Bj|A) = mj / P(A).
Common checks
Posteriors across all cases sum to 1.
Units: percents vs decimals (be consistent).
Every case must have a prior and a likelihood.
Part B: Permutations & Combinations
Formulas
Permutations: P(n,r) = n! / (n − r)!
Combinations: C(n,r) = n! / [r!(n − r)!]
When to use
Roles/arrangements/schedules → permutations.
Groups/committees/selections → combinations.
Replace vs. without replacement: affects counts.
Part C: Discrete Models
Poisson Distribution
Probability mass function: P(X = k) = (e−λ λk) / k!
Mean parameter: λ = (rate) × (interval)
Shortcut: At least one event = 1 − P(0) = 1 − e−λ
Binomial Distribution
Probability mass function: P(X = k) = C(n,k) pk (1 − p)n−k
Use for: exact successes, cumulative (≤ or ≥), or complements.
Geometric Distribution
Probability mass function: P(X = k) = (1 − p)k−1 p
Represents the trial number of the first success.
Support: k = 1, 2, 3, …
Parameter hygiene: Define n, p, λ before you compute. Convert units (minutes → hours) when building λ. Use complements to simplify cumulative probabilities.
Smart Exam Tips
Underline what’s asked: exactly k, at least, at most, or a model name.
For model ID, rule out the other three with a one-line reason.
Check sums: posteriors sum to 1; binomial probabilities over all k sum to 1.
Use the complement trick for “at least one”.
Keep arithmetic symbolic as long as possible; round at the end.
Quick Checklist
Can I explain independence vs. mutual exclusivity with an example?
Do I know when to use permutations vs. combinations?
Can I set up and solve a Bayes table with 3–5 categories?
Given a short scenario, can I pick Binomial / Poisson / Geometric / Uniform and justify why?
Do I remember the core formulas for Poisson, Binomial, and Geometric and how to do “at least/at most”?
Good luck! Master the patterns above and you’ll be ready for both the conceptual and calculation parts.