What is the difference between poisson, t, chi-squared, binomial, and normal distributions?

The core distinction between the Poisson, t, chi-squared, binomial, and normal distributions lies in the specific type of data and underlying random process each is designed to model, which in turn dictates their mathematical properties and primary applications in statistical inference. The normal distribution is a continuous, symmetric probability distribution defined by its mean and variance, famously forming a bell curve; it is foundational due to the Central Limit Theorem, which states that means of large samples from any population tend toward normality, making it the cornerstone for confidence intervals and hypothesis tests concerning means. The binomial distribution is discrete, modeling the number of successes in a fixed number of independent trials, each with the same probability of success, such as counting defective items in a batch or the number of heads in coin flips. The Poisson distribution is also discrete, but it models the number of events occurring in a fixed interval of time or space when these events happen with a known constant mean rate and independently of the time since the last event, making it ideal for counts like customer arrivals or radioactive decay events.

The t and chi-squared distributions are fundamentally sampling distributions, derived from the normal distribution and critical for inference. The chi-squared distribution arises as the sum of the squares of independent standard normal variables; it is skewed to the right and is pivotal in tests for categorical data goodness-of-fit and independence, as well as in forming confidence intervals for a population variance. The t-distribution, similar in shape to the normal but with heavier tails, describes the distribution of a sample mean when the population standard deviation is unknown and is estimated from the sample; it is therefore the exact sampling distribution for means under normality when using the sample standard deviation, making it indispensable for small-sample inference about a population mean. The heavier tails of the t-distribution account for the extra uncertainty introduced by estimating the population variance, and it converges to the normal distribution as the sample size grows.

The practical implications of these differences are profound for correct analytical methodology. Using a binomial model for a low-probability event over many trials might be approximated by a Poisson, but applying a normal approximation to a binomial count requires a sufficiently large sample size to avoid error. Misapplying the normal distribution to construct a confidence interval for a mean with a small sample and unknown variance would underestimate uncertainty, whereas correctly using the t-distribution provides an accurate interval. Similarly, using a normal distribution to model count data can lead to invalid predictions and probabilities below zero, while a Poisson or binomial model respects the discrete, non-negative nature of the data. The chi-squared distribution's role is distinct, as it is not typically a model for raw data but a tool for evaluating hypotheses about variance or categorical distributions, where its asymmetry is a key feature.

Ultimately, the choice among these distributions is not interchangeable but is dictated by the data-generating mechanism, the scale of the data (continuous vs. discrete count), and the specific inferential goal (e.g., estimating a proportion, a mean, a variance, or testing for independence). The normal distribution's centrality stems from asymptotic theory and its mathematical tractability, while the binomial and Poisson address fundamental discrete random processes. The t and chi-squared distributions are specialized tools for inference, extending the utility of the normal model to realistic scenarios where parameters like variance are unknown or when dealing with quadratic forms of normal variables. Understanding their distinct origins—from counting successes and rare events to describing the behavior of sample statistics—is essential for valid statistical modeling and hypothesis testing.