Modeling Maximum Extreme Values by New Different Continuous Distributions Derived from the Median Based Unit Rayleigh Distribution

Imagine trying to predict the worst possible event a system might ever face: the highest flood level a river could reach in a century, the most intense heatwave a city might endure, or the strongest wind a bridge must withstand. These are not just random numbers; they are the boundaries of safety and survival. For decades, scientists have relied on a specific mathematical framework to model these extremes, a framework that suggests that no matter how complex the underlying data is, the very largest values tend to settle into one of three predictable patterns as the sample size grows infinitely large. This is the realm of extreme value theory, a field dedicated to understanding the behavior of the most rare and dangerous occurrences. While the theory tells us what happens when we wait forever for the ultimate record, the real world rarely offers infinite time. We work with finite data, and in that limited window, the path to the extreme can be treacherous and misleading if we use the wrong tools.

In a recent study, researcher Iman Attia set out to build better tools for this specific challenge. The work begins with a known statistical shape called the Median Based Unit Rayleigh distribution, a model that describes data confined between zero and one. Attia asked a fundamental question: if we start with this specific shape and apply different mathematical transformations to it, what new shapes emerge, and how do they behave when we look at their most extreme values? The goal was not just to create new formulas, but to engineer distributions that fit real-world data more accurately before the sample size becomes infinite. By taking the original shape and twisting it through specific operations—such as taking the negative logarithm of the variable or applying a double logarithmic transformation—the author derived four distinct new continuous distributions. Each of these new models comes with its own unique set of properties, including how likely it is to produce a value at a certain point, how that likelihood accumulates, and how the risk of an extreme event changes as time or magnitude increases.

The study reveals that these new distributions are not just theoretical curiosities; they possess a powerful ability to adapt to the data they are meant to describe. When the author analyzed the behavior of the maximum values within these new models as the number of observations grew, a clear pattern emerged. Depending on which transformation was applied, the models converged toward one of two major types of extreme value behavior. Two of the new distributions, created by applying a negative log squared transformation and a negative double log transformation, were shown to eventually settle into the Gumbel pattern. This pattern is characteristic of light-tailed distributions, where extreme values are rare and the probability of them occurring drops off quickly. Another transformation, involving a specific power-law adjustment, led the model to converge toward the Fréchet pattern, which describes heavy-tailed distributions where extreme events are more frequent and the tail of the probability curve is much longer.

Perhaps most significantly, the study confirmed that the original Median Based Unit Rayleigh distribution itself, without any transformation, converges to a third pattern known as the Weibull distribution. This pattern is unique because it applies to data that has a hard upper limit, a maximum ceiling that cannot be exceeded. By proving these limits, the author demonstrated that these new models are mathematically sound and belong to the correct families of extreme value theory. The research goes beyond just identifying the limits; it provides the complete mathematical machinery for each new distribution. This includes the ability to calculate the average value, the variance, and the specific point at which a certain percentage of data falls below. It also details the "hazard rate," a measure of how the risk of an extreme event changes over time, showing that for some of these new models, the risk starts low, rises to a peak, and then stabilizes or drops, offering a more nuanced view of danger than simpler models.

The value of this work lies in its precision for finite data. While the ultimate limit of any extreme value model might be a standard Gumbel or Fréchet distribution, the journey there matters immensely. Standard models often assume a rigid speed of convergence that does not match reality, leading to biased estimates when applied to the finite datasets we actually possess. These new distributions act as a bridge. They contain extra parameters that allow them to bend and shape themselves to fit the specific characteristics of the data at hand, capturing the "penultimate" behavior of the system before it reaches the theoretical infinity. This means that engineers, hydrologists, and risk analysts can use these models to get a much more accurate picture of potential disasters. They can trust that the model understands the physical or structural nuances of the system they are studying, rather than just forcing the data into a generic box. The study concludes that by deriving these specific transformations from a single parent distribution, we gain a versatile toolkit for modeling maximum extreme values, ensuring that our predictions for the worst-case scenarios are as reliable and grounded in reality as possible.