📊 statistics

A New Generated Class: Theory and Inference with Application to Daily COVID-19 Cases in Region Aseer, Saudi Arabia

This paper introduces a flexible trigonometric-generated family of distributions, specifically a parsimonious NewWeibull special case, which is mathematically validated through theoretical derivations and Monte Carlo simulations to effectively model heterogeneous outbreak data, outperforming existing models when applied to daily COVID-19 cases in Saudi Arabia's Aseer Region.

Original authors: Muhammad Arshad, Ali Algarni

Published 2026-08-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Muhammad Arshad, Ali Algarni

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Tracking the daily rise and fall of infectious disease cases is a task that looks simple on a map but is mathematically treacherous in practice. When an outbreak begins, the number of new cases rarely follows a smooth, predictable line. Instead, the data often arrives in jagged bursts, with sudden spikes that shoot far above the average and long, heavy tails where cases linger at low levels for weeks. Standard statistical tools, which were designed to describe more orderly phenomena like human height or test scores, often fail to capture this erratic behavior. They struggle to model the extreme volatility seen in real-world epidemics, leaving public health officials without a precise way to understand the risk or predict the next surge. To solve this, researchers need flexible mathematical models that can bend to fit the chaotic shape of actual outbreak data, rather than forcing the data to fit a rigid, pre-existing shape.

In a recent study focused on the Aseer region of Saudi Arabia, a team of mathematicians developed a new tool specifically designed to handle this kind of messy, real-world data. They created a new family of probability distributions, which are essentially mathematical recipes for describing how likely different outcomes are. Their specific creation, called the NewWeibull distribution, builds upon a classic model known as the Weibull distribution but adds a layer of trigonometric flexibility. Think of the original model as a standard ruler that works well for straight lines, while this new version is like a flexible measuring tape that can curve to fit complex, irregular shapes without losing its accuracy. The researchers did not just propose this new formula; they spent time proving it works mathematically, testing it with thousands of computer simulations, and finally applying it to real daily case counts from five different cities in Saudi Arabia to see if it could outperform the established methods used by epidemiologists.

The core of the study involved taking the daily counts of COVID-19 cases from five distinct urban centers in the Aseer region: Ahad Rifaydah, Billasmar, Sarat Abidah, Al Harajah, and Abha. Each city presented a unique challenge. Some, like Billasmar, showed extreme volatility with rare but massive spikes in cases, while others, like Abha, displayed more consistent, high-level transmission with fewer wild swings. The researchers fed this data into their new NewWeibull model and compared its performance against a lineup of traditional models, including the standard Weibull, Gamma, and LogNormal distributions. They measured success by checking how closely the model's predictions matched the actual data, paying special attention to how well it captured the extreme tails of the distribution—the rare, high-risk days that are often the most critical for planning.

The results were clear and consistent across all five cities. The NewWeibull distribution consistently provided a better fit to the data than any of the traditional models. In statistical terms, it achieved lower error scores for how well it described the overall shape of the data and, crucially, how well it captured the heavy tails where the most dangerous spikes occurred. For instance, in the city of Abha, the new model reduced the error in describing the tail of the data by about six percent compared to its closest competitor. In Ahad Rifaydah, it achieved a near-perfect match with the actual data, while older models either missed the peaks entirely or failed to account for the long tail of low-level cases. The study also noted that some of the more complex, existing flexible models actually performed worse, suggesting that adding too many moving parts can sometimes make a model less reliable rather than more.

To ensure these findings were not just a lucky coincidence with this specific dataset, the authors subjected their new model to rigorous testing using computer simulations. They generated thousands of fake datasets that mimicked the behavior of real outbreaks and tried to estimate the model's parameters from these synthetic numbers. The simulations showed that the method used to find the best-fitting numbers for the model was highly reliable. As the amount of data increased, the estimates became more precise and the errors shrank at the expected rate. The researchers found that while the scale of the outbreak was easy to estimate even with small amounts of data, the shape of the curve required larger datasets to pin down accurately. This gave them a practical guideline: for reliable predictions about the volatility of an outbreak, health officials should wait until they have at least several hundred days of data before trusting the model's specific shape parameters.

The study concludes that this new mathematical approach offers a powerful, yet simple, way to understand the unpredictable nature of disease spread. By combining a solid theoretical foundation with real-world validation, the NewWeibull distribution provides a tool that is both mathematically sound and practically useful. It does not require complex, ad-hoc adjustments to fit the data; instead, it naturally accommodates the heavy tails and right-skewed patterns that characterize epidemics. For public health teams in regions with diverse geography and population densities, like the mountainous and varied terrain of the Aseer region, having a model that can accurately describe both the calm periods and the sudden surges is essential. The work demonstrates that with the right mathematical framework, it is possible to turn chaotic, jagged outbreak data into a clear, understandable picture of risk, helping officials make better decisions about where to allocate resources and how to prepare for the next wave.

Technical Summary: A New Generated Class: Theory and Inference with Application to Daily COVID-19 Cases in Region Aseer, Saudi Arabia

Problem Statement
Applied statistics faces a persistent challenge in modeling heterogeneous data characterized by irregular behavior, extreme values, and heavy tails. Classical distributions (e.g., Weibull, Gamma, LogNormal) often fail to capture the specific features of real-world datasets, particularly in fields like reliability engineering, survival analysis, and epidemiology. The authors note that while heavy-tailed (HT) models are superior for such anomalies, few probability models naturally embody these properties while maintaining analytical tractability. This gap is particularly evident in modeling localized COVID-19 outbreaks in regions with complex topography and uneven population density, such as the Aseer Region of Saudi Arabia, where daily case data exhibit strong right-skew, heavy tails, and shifting hazard shapes that strain classical assumptions.

Methodology
To address these limitations, the paper proposes a Trigonometric-Generated (T–G) family of distributions. The core innovation is a new generator function defined by the cumulative distribution function (CDF):
$F_{X}(x; \theta) = \frac{2 G(x; \theta)}{1 + \sin(\frac{\pi}{2} G(x; \theta))}$
where $G(x; \theta)$ is the baseline CDF. The corresponding probability density function (PDF), survival function, hazard rate, and cumulative hazard are derived analytically.

The study focuses on a parsimonious special case of this family: the NewWeibull distribution, obtained by using the standard Weibull distribution as the baseline. The methodology includes:

Theoretical Derivation: Closed-form expressions for raw moments are derived using a Maclaurin expansion of the generator function $H'(u)$ , establishing conditions for valid series–integral interchange. The paper also derives expressions for Shannon Entropy, order statistics, incomplete moments, and various inequality curves (Lorenz, Bonferroni, Zenga).
Inference: Maximum Likelihood Estimation (MLE) is employed. The paper establishes regularity conditions to ensure the identifiability of the shape ( $k$ ) and scale ( $\eta$ ) parameters and proves the asymptotic normality of the estimators.
Simulation: A Monte Carlo study evaluates the finite-sample performance of the MLEs across sample sizes $n = 50$ to $2000$. Metrics include bias, standard deviation, and Mean Squared Error (MSE).
Empirical Application: The model is applied to daily COVID-19 case data from five urban centers in the Aseer Region (Ahad Rifaydah, Billasmar, Sarat Abidah, Al Harajah, and Abha). Performance is compared against classical models (Weibull, Gamma, LogNormal, Exponential) and alternative generated families (Exponentiated-Exponential, SinWeibull) using Kolmogorov–Smirnov (KS) statistics, Anderson–Darling (ADE), and Cramér–von Mises (CVME) criteria.

Key Contributions

New Distribution Family: Introduction of the T–G family, which offers a flexible framework for generating new distributions without introducing extra parameters beyond the baseline.
Analytical Tractability: The NewWeibull distribution provides closed-form expressions for moments, entropy, and order statistics, facilitating rigorous statistical inference.
Theoretical Guarantees: The paper provides formal proofs for the existence of moments, the validity of the series expansion, and the asymptotic properties of the MLEs.
Empirical Superiority: The NewWeibull distribution demonstrates consistent statistical superiority over competing models in fitting heterogeneous outbreak data, particularly in capturing tail behavior and global fit simultaneously.

Results

Statistical Properties: Numerical analysis reveals that the scale parameter $\eta$ controls the magnitude of moments, while the shape parameter $k$ governs tail weight and asymmetry. Low $k$ values produce heavy tails and high skewness, while high $k$ values lead to lighter tails and mesokurtic behavior.
Simulation Performance: The MLEs for both parameters are consistent and asymptotically normal. The scale estimator ( $\hat{\eta}$ ) is effectively unbiased even at small sample sizes ( $n=50$ ). The shape estimator ( $\hat{k}$ ) exhibits a small positive bias at very small $n$ (e.g., bias $\approx 0.09$ at $n=50$ for $k=3$ ), which dissipates rapidly as $n$ increases, becoming negligible by $n \approx 1000$ . Mean Squared Errors decrease monotonically with sample size.
Epidemiological Application: In all five cities analyzed, the NewWeibull distribution achieved the lowest Anderson–Darling and Cramér–von Mises statistics while maintaining high KS p-values.
- In Ahad Rifaydah, it achieved a CVME of 0.028 and a KS p-value of 0.989, significantly outperforming the SinWeibull model, which showed signs of overfitting (ADE > 4.5).
- In Billasmar (characterized by heavy tails), the NewWeibull provided the most balanced fit, whereas the Exponential and SinWeibull models degraded significantly.
- The model successfully adapted to diverse outbreak profiles, from concentrated urban transmission (Abha) to volatile, low-density scenarios (Billasmar).

Significance and Claims
The paper claims that the NewWeibull distribution represents a mathematically sound, simulation-validated, and empirically effective tool for outbreak modeling. Its significance lies in its ability to bridge the gap between analytical tractability and the flexibility required to model heavy-tailed, skewed data without the instability observed in more complex generated families (like the SinWeibull in this study).

The authors position the T–G framework as a parsimonious approach that preserves closed-form solutions while supporting rigorous inference. They conclude that the model is particularly valuable for underrepresented regions with complex geographic and demographic factors, offering a practical component for modern epidemiological analytics. The study emphasizes that while the model is robust, it is intended for marginal distributions of daily counts and should be accompanied by standard diagnostics for temporal dependence and reporting noise.

More like this