Probabilistic foundations of Binomial & Hypergeometric Sampling.
The Bernoulli process represents the quintessential mathematical model of discrete randomness, consisting of repeated independent binary trials with invariant success probability. Aggregating counts across fixed sample sizes gives rise to the Binomial distribution, while drawing without replacement from finite structures introduces hypergeometric dependence. In this laboratory, we investigate the fundamental variance scaling of discrete processes, contrast infinite-population replacement models with finite-population urn sampling, and quantitatively confirm the single-trial Bernoulli variance parameter m = 0.240. In mathematical statistics, empirical data science, and modern computational engineering, understanding the asymptotic sampling properties and theoretical behavior of Bernoulli Process, Binomial Distribution, and Hypergeometric Sampling provides crucial guarantees. Rigorous experimental validation connects foundational probability laws with observable empirical phenomena, ensuring that statistical inference remains valid under finite sample sizes and real-world data collection constraints.
Viva voce questions & analytical theory
How is a Bernoulli trial defined, and how does it generate the Binomial distribution?
A Bernoulli trial is a random experiment with exactly two mutually exclusive outcomes: 'success' (X = 1) with probability p and 'failure' (X = 0) with probability q = 1 - p. The sum of n independent, identically distributed Bernoulli trials X = ∑_{i=1}^n X_i constitutes a Binomial random variable X ~ Binomial(n, p). Its probability mass function is P(X = k) = (n choose k) p^k (1-p)^(n-k) for k ∈ {0, 1, ..., n}. By linearity of expectation, E[X] = n E[X_1] = np, and because the trials are mutually independent, Var(X) = n Var(X_1) = np(1-p).
What is the key difference between Binomial and Hypergeometric sampling models?
The Binomial distribution models sampling with replacement (or sampling from an infinite population), where the probability of success p remains strictly invariant across successive draws, preserving trial independence. The Hypergeometric distribution models sampling without replacement from a finite population of size N containing M successes and N - M failures. In hypergeometric sampling, successive draws are statistically dependent; removing an item alters the composition of the remaining pool.
What is the Finite Population Correction (FPC) factor, and when is it negligible?
The variance of a Hypergeometric random variable is Var(X) = n (M/N) (1 - M/N) [(N - n)/(N - 1)]. The term (N - n)/(N - 1) is the Finite Population Correction (FPC) factor. Because sampling without replacement prevents re-drawing previously selected items, it yields greater information per sample, compressing variance relative to the binomial case. When the sampling fraction n/N < 0.05, the FPC factor is greater than 0.95 and is conventionally deemed negligible, allowing binomial approximation.
Under what conditions does the Binomial distribution converge to the Poisson distribution?
According to the Law of Rare Events (Poisson Limit Theorem), if the number of trials n → ∞ and the success probability p → 0 such that the expected value λ = np remains constant, the Binomial PMF converges pointwise to the Poisson PMF: lim_{n → ∞} (n choose k) (λ/n)^k (1 - λ/n)^(n-k) = (e^(-λ) λ^k) / k!. The total variation distance between the two distributions is bounded by ∑ p_i^2, ensuring high fidelity when p ≤ 0.05.
How does the De Moivre–Laplace theorem bridge Binomial and Normal distributions?
The De Moivre–Laplace theorem is a special case of the Central Limit Theorem. As n → ∞ with fixed p, the standardized binomial variable Z = (X - np) / √(np(1-p)) converges in distribution to standard normal N(0, 1). In practical applications, the normal approximation is deemed accurate when both np ≥ 10 and n(1-p) ≥ 10, with a continuity correction of ±0.5 applied to account for the discrete-to-continuous domain translation.
Why is the Bernoulli variance maximized at p = 0.5?
The Bernoulli variance function is f(p) = p(1-p) = p - p^2. Differentiating with respect to p yields f'(p) = 1 - 2p. Setting the derivative to zero yields the critical point p = 0.5. The second derivative f''(p) = -2 < 0 confirms a global maximum at p = 0.5, where Var(X) = 0.25. When p approaches 0 or 1, the outcome becomes nearly deterministic and variance collapses toward zero.
What role does the Negative Hypergeometric distribution play in sequential sampling?
The Negative Hypergeometric distribution models the number of draws without replacement required from a finite population to observe exactly r successes. It is the finite-population analogue of the Negative Binomial distribution and finds extensive application in industrial acceptance sampling, ecological capture-recapture surveys, and lottery audit verification.
Experimental precautions and statistical checks
- Ensure the pseudorandom generator seed produces strictly uncorrelated sequences across consecutive trial blocks.
- Apply a half-integer continuity correction (±0.5) when approximating binomial cumulative sums with continuous Gaussian integrals.
- Verify that population size N exceeds sample size n in hypergeometric modes to prevent empty-urn sampling errors.
- Monitor floating-point precision when computing high-order binomial coefficients (n choose k) to avoid arithmetic overflow.
Theoretical assumptions & limitations
- Binomial models assume strict outcome independence, which fails in contagious or clustered event series.
- Normal approximations to the binomial distribution deteriorate significantly when p is extreme (p < 0.05 or p > 0.95).