The binomial distribution is useful as a model for real-world processes where there are fixed numbers of trials, each with two possible outcomes (success or failure), such as quality control testing, clinical trials, and election predictions . It requires independence between trials and a constant probability of success across the trials. For example, in manufacturing, modeling the number of defective products allows for planning and quality assurance interventions. However, these applications are constrained when dependencies or varying probabilities are present, thus requiring consideration or alternative models like the hypergeometric or negative binomial distribution . Understanding these constraints ensures a reliable application of the binomial model across various domains.

To calculate the probability of exactly k successes in a hypergeometric distribution, use the hypergeometric probability formula: P(Y=k) = [C(M, k) × C(N-M, n-k)] / C(N, n), where C(a, b) represents combinations of a taken b at a time, M is the number of successes in the population, N is the total population size, and n is the sample size . This formula calculates the probability by considering the combinations of selecting k from the successes and n-k from the failures, divided by the total sample combinations from the population.

The binomial distribution is often used as an approximation for hypergeometric distributions when the sample size n is small relative to the population N, making the ratio n/N negligible . This approximation is valid because under these conditions, the probability of success remains nearly constant regardless of the outcome of previous draws (approximating independence), allowing the hypergeometric distribution's behavior to resemble that of the binomial distribution . The larger the population, the more accurate the approximation because the dependence introduced by sampling without replacement becomes minimal.

For a model using stock prices to be appropriately described by a binomial distribution, the daily returns must be independent repetitions of the experiment . This assumes that the daily returns are independent and identically distributed, each with a probability p of a positive return and 1-p for a negative return. However, this assumption is contentious, as stock prices can be influenced by numerous factors leading to dependency in returns. The validity of modeling stock prices binomially rests on the assumption that these dependencies are negligible or accounted for in the modeling . Thus, discerning the independence and consistency of stock price changes across observation periods is crucial.

The expected value and variance are fundamental in determining the probability of outcomes in a statistical experiment. The expected value, or mean, indicates the central tendency or average number of successes in experiments under repetition. Variance quantifies the spread or dispersion of the probability distribution, indicating how much the actual results may vary from the expected value . Together, they define the confidence and reliability of different outcomes occurring by illustrating the distribution's central spine and variability around it, enabling analysts to assess odds and make informed predictions based on distribution patterns.

The concept of independence is central to determining whether a statistical experiment is binomial. For an experiment to be binomial, each trial must be independent, meaning the outcome of one trial does not affect the outcome of another. For example, tossing a die multiple times is binomial because each toss is independent. Conversely, drawing samples without replacement from a small population, such as TVs, is not binomial because the samples affect each other's outcomes, violating the independence requirement . Independence ensures that trials do not influence each other, which is fundamental for modeling using a binomial distribution.

The variance of a binomial distribution with parameters n = 100 and p = 0.35 is calculated as V(Y) = np(1-p) = 100 × 0.35 × (1-0.35) = 22.75 . For a hypergeometric distribution with n = 100, M = 350, and N = 1000, the variance is V(Y) = np(1-p) × (N-n)/(N-1) = 100 × (350/1000) × ((1000-350)/1000) × 900/999 = 20.4955 . The variance of the binomial distribution is larger than that of the hypergeometric distribution due to the correction factor (N-n)/(N-1) in the hypergeometric distribution that reduces the variance.

Both hypergeometric and binomial distributions have the same expectation formula E(Y) = np because they both model scenarios involving a fixed number of successes in a series of trials . However, they differ in their variance calculations due to the nature of sampling. The binomial distribution assumes independence between trials, thus its variance is V(Y) = np(1-p). However, in a hypergeometric distribution, the lack of replacement in sampling introduces dependency between trials, which is adjusted with a finite population correction factor, resulting in V(Y) = np(1-p) × (N-n)/(N-1). This accounts for the reduction in variability because sampling without replacement ties the outcomes together.

The variance of a hypergeometric distribution includes a finite population correction factor calculated as (N-n)/(N-1), which adjusts the variance downwards . This adjustment is necessary because hypergeometric sampling is done without replacement, leading to dependence among trials. The correction factor accounts for the decreasing population size after each draw, which ties the outcomes more closely together. This contrasts with the binomial variance which does not require such an adjustment as the trials are independent, due to sampling with replacement or the assumption thereof . The correction ensures the variance accurately reflects the reduced variability due to sampling dependencies.

A Poisson distribution is used as an approximation for a binomial distribution when the number of trials n is large, and the probability of success p is small such that np is moderate (often approaching values around 10 or less). Under these circumstances, the binomial distribution’s skewness and shape converge to that of a Poisson distribution as the probability of any individual event occurring becomes low, but collectively these rare events produced an expected result . This approximation is particularly useful for simplifying complex binomial calculations in large-sample low-probability scenarios.