28.3. Distributions of Random Variables

Distribution of variables can be mathematically represented.

This statement introduces the concept that distributions don't just exist in a theoretical sense; they can also be expressed through mathematical formulas. This means we can quantify how likely different outcomes are, which is crucial in probability and statistics.

Think of it like setting up a scoring system for a game. Instead of just saying a player can score between 0 to 100 points, we can specify how likely they are to score 90 points versus 70 points using a distribution.

Uniform distribution implies that any value between x_min and x_max is equally likely to occur.

In a uniform distribution, every outcome within the specified range has the same chance of happening. This is like rolling a fair die where each number from 1 to 6 has an equal probability of appearing. Uniform distributions are easy to understand and often used in situations where no outcome is favored over another.

Imagine you are picking a random day in a month. If the month has 30 days, each day has an equal likelihood of being chosen, similar to drawing numbers from a hat where each number has a fair chance.

The general normal (or Gauss) distribution is given by, Fig. 28.1: 1 1[x−μ]² 𝓕(x) = e^(−2σ²) 2πσ A normal distribution N(μ,σ²) can be normalized by defining x−μ y = (28.11) σ and y would have a distribution N(0,1): 1 y² 𝓕(y) = e^(−2) p2π

Normal distributions exhibit a symmetrical bell-shaped curve where values cluster around a mean (μ). The spread is determined by the standard deviation (σ). This distribution is foundational in statistics because many natural phenomena fall under this category, reflecting randomness surrounded by average behavior.

Imagine the height of adult males in a country. Most men will be around the average height, with fewer men being extremely tall or very short. When plotted, this data would generally form a bell-shaped curve, which is exactly what a normal distribution describes.

A random variable is lognormally distributed if the natural logarithm of the variable is normally distributed.

A lognormal distribution arises when data values cannot be negative and tend to cluster around a minimum value but can increase significantly. This is common in scenarios like income or stock prices, where most values are low, but a few can be extremely high.

Think of real estate prices in a city; while most homes are priced similarly within a range, a few properties, such as luxury estates, can skyrocket to millions of dollars, thus creating a lognormal shape when graphed.

Beta distributions are very flexible and can assume a variety of shapes including the normal and uniform distributions as special cases. On the other hand, the beta distribution requires four parameters.

Beta distributions are versatile because they can take on various forms — from uniform to bell-shaped. This makes them suitable for modeling probabilities in specific ranges, particularly where we are interested in proportions or probabilities between 0 and 1.

If you're looking at the probability of success in a project that could either succeed brilliantly or fail, a beta distribution can effectively model the chances of varying levels of success, depending on how many favorable conditions are met.

BiNormal distribution

A BiNormal (or bivariate normal) distribution involves two variables where each is normally distributed. The relationship between the two is often illustrated in a three-dimensional space, helping to visualize how one variable influences another.

In science, when examining the relationship between test scores and study time, you might find that increasing study time usually increases scores. When plotted together, this could form a BiNormal distribution.

--