In a theoretical study published in the Journal of Statistical Theory and Applications, researchers Manal Alharbi, Abdullah Almarashi, and Muhammad Qaiser Shahbaz introduced two extensions of the log-logistic distribution: the generalized order statistics transmuted log-logistic distribution and the dual generalized order statistics transmuted log-logistic distribution. The authors developed these probability models to address persistent modeling challenges in statistical theory where traditional distributions fail to adequately capture complex real-world data structures. As empirical phenomena in applied sciences grow increasingly diverse and skewed, conventional distribution models often lack the mathematical flexibility needed to fit observations exhibiting heavy tails or intricate variation. By embedding generalized order statistics and dual generalized order statistics into transmuted distribution families, the research team established adaptable statistical frameworks capable of handling censored and survival datasets. Their theoretical investigation delivers complete mathematical formulations for distribution functions, densities, reliability characteristics, hazard rates, moments, and parameter estimation.
The baseline model underlying both constructions is the classical log-logistic distribution, also known as the Fiske distribution in economic literature focusing on income distributions. Governed by positive shape and scale parameters, the log-logistic distribution has extensive utility across reliability engineering and mathematical finance, where failure times and heavy-tailed observations generate pronounced fluctuations. The authors noted that the model has established applications in supply chain systems, including stochastic modeling for warehouse location and retailer allocation, as well as joint modeling in environmental and forestry contexts using copula-based methods. While the cumulative distribution function of the log-logistic model provides a valuable analytical basis, its classical structure remains restricted when confronting subtle variation in empirical data. By modifying this baseline through additional parameters, statisticians seek to expand model adaptability without sacrificing analytical tractability, enabling more nuanced interpretations of complex operational failure times and survival patterns.
To provide theoretical context, the authors reviewed previous methodologies used to extend classical distributions through transmutation techniques. Shaw and Buckley initially introduced the quadratic rank transmuted distribution, which incorporates a bounded transmutation parameter into cumulative distribution functions to generate quadratic flexibility. Subsequent studies expanded this approach into cubic transmutation models utilizing the first three order statistics, as well as record-value frameworks utilizing upper and lower record statistics. More recently, researchers introduced transmuted distribution families based on generalized order statistics and dual generalized order statistics, leveraging their first two ordered elements. By synthesizing the log-logistic baseline with these generalized and dual generalized order statistics transmutation mechanisms, Alharbi, Almarashi, and Shahbaz created two extended models designed to interpret subtle differences across diverse empirical distributions.
The first model presented in the study is the generalized order statistics transmuted log-logistic distribution, characterized by five parameters: shape parameter beta, positive scale parameter theta, transmutation parameter lambda bounded between negative one and positive one, and generalized order statistics parameters k and m, with k at least one and m any real number other than negative one. In Theorem 1, the authors formally established its cumulative distribution function, proving the expression by inserting the log-logistic baseline into the transmuted generalized order statistics family. They then derived the probability density function in Definition 1, illustrating that the density function is structured around weighted rational powers of the baseline distribution. Graphical plots of the distribution’s density and cumulative functions across varied parameter values demonstrated notable shifts in skewness, peak height, and tail weight, confirming that the new distribution can adapt to varied observational shapes.
The authors subsequently analyzed the reliability properties and hazard rate behavior of the generalized order statistics transmuted log-logistic distribution. They derived closed-form mathematical equations for both the reliability function and the hazard rate function, highlighting that the model’s hazard rate is expressed as the baseline log-logistic hazard rate multiplied by a rational term dependent on the transmutation and order parameters. Visual representations in the paper revealed that selecting different parameter combinations produces diverse hazard rate trajectories, accommodating varied operational lifetime patterns. Furthermore, in Theorem 2, the authors obtained the general r-th moment of the distribution by computing probability-weighted moments of the log-logistic baseline, expressing the final formula in terms of complete beta functions. From this formulation, Definition 2 established the explicit population mean of the distribution, while the paper also detailed the moment-generating function as an infinite series weighted by factorials and beta functions, alongside the quantile function equation.
To assess informational complexity, the researchers evaluated Rényi entropy for the generalized order statistics transmuted log-logistic distribution. In Theorem 3, they formulated an explicit analytical expression for Rényi entropy, proving the result by expanding powered density terms using the binomial theorem and resolving the integral via probability-weighted moments into a summation involving complete beta functions. Moving to the second proposed model, the authors introduced the dual generalized order statistics transmuted log-logistic distribution in Section 4. Theorem 4 detailed its cumulative distribution function, built upon dual generalized order statistics with parameters shape beta, scale theta, bounded transmutation parameter lambda, and dual parameters k and m. Definition 3 provided the corresponding probability density function.
Analytical properties of the dual generalized order statistics transmuted log-logistic distribution were derived in parallel detail. The researchers established formulas for the dual reliability function and hazard rate function, showing that the hazard rate reflects the baseline log-logistic density scaled by a rational expression incorporating dual parameters. In Theorem 5, the authors obtained the r-th raw moment through probability-weighted moments of the baseline distribution, expressing the outcome using complete beta functions. Definition 4 supplied the explicit equation for the distribution mean, while subsequent equations outlined the moment-generating function and the non-linear quantile equation. Furthermore, Theorem 6 established the derivation of Rényi entropy for the dual distribution, employing binomial expansion of the powered density terms to yield an analytical expression featuring beta functions. These mathematical derivations confirm that the dual model retains the analytical completeness of the generalized distribution while offering distinct structural flexibility for empirical modeling.
To facilitate empirical applications, the authors formulated maximum likelihood estimation frameworks for both proposed distributions across a five-parameter vector comprising lambda, beta, theta, k, and m. For each model, the complete sample log-likelihood function was defined based on independent random samples of size n. The authors derived the five elements of the score function vector by calculating partial derivatives of the log-likelihood with respect to each individual parameter, explicitly defining auxiliary functions W1 and W2 in the process. Setting these score equations simultaneously to zero yields the maximum likelihood estimators for the underlying parameters. Because the resulting non-linear systems lack analytical closed-form solutions, the paper explained that parameter estimation must be conducted numerically. Practitioners can solve the score equations simultaneously using standard software packages or iterative numerical algorithms such as the multi-dimensional Newton-Raphson method, which iteratively updates parameter estimates until convergence criteria are satisfied across observed datasets.
The authors validated the consistency and numerical performance of their maximum likelihood estimators through extensive simulation studies. For the generalized order statistics transmuted log-logistic distribution, random samples of sizes 50, 100, 500, and 1000 were simulated using fixed parameter values: beta equal to 3, theta equal to 2, lambda equal to 0.5, k equal to 5, and m equal to 6.5. For the dual generalized order statistics distribution, samples of sizes 50, 75, 150, and 300 were generated under the same baseline parameter values. In both simulation series, parameters were estimated across 10,000 independent replications to compute average parameter estimates alongside mean square errors. The reported results in Tables 1 and 2 showed that average parameter estimates closely aligned with true parameter values across all evaluated configurations. Furthermore, estimated mean square errors exhibited consistent decreases as sample sizes increased, demonstrating the numerical stability, statistical consistency, and practical effectiveness of the maximum likelihood estimation technique for both extended distributions.
Finally, the research team demonstrated the practical effectiveness of both distributions using two real-world empirical datasets from the United States. The first dataset examined the logarithm of shares offered by 482 American corporations, while the second analyzed the percentage of population living below the poverty line across 440 American counties. The proposed models were evaluated alongside competing models, including the classical log-logistic, quadratic transmuted log-logistic, and record-based transmuted log-logistic distributions. Model comparisons utilized goodness-of-fit criteria: maximized log-likelihood, Akaike’s information criterion, corrected Akaike’s information criterion, Bayesian information criterion, and Kolmogorov-Smirnov, Anderson-Darling, and Cramér-von Mises tests. Goodness-of-fit statistics yielded high p-values exceeding the 0.05 threshold for both proposed models. In addition, likelihood ratio tests and Akaike weights showed that the generalized and dual generalized order statistics transmuted log-logistic distributions achieved significantly higher weights and superior fit compared to competing distributions, confirming their value for complex empirical modeling.
Subject of Research: Statistical distribution theory and transmuted log-logistic modeling
Article Title: New Generalized and Dual Generalized Order Statistics based Transmuted Log-Logistic Distributions
Article References: Alharbi, M., Almarashi, A., & Shahbaz, M. Q. (2026). New Generalized and Dual Generalized Order Statistics based Transmuted Log-Logistic Distributions. Journal of Statistical Theory and Applications, 25(1), Article 1. https://doi.org/10.1007/s44199-025-00152-9
Image Credits: AI Generated
DOI: 10.1007/s44199-025-00152-9
Keywords: log-logistic distribution, generalized order statistics, dual generalized order statistics, transmuted distributions, heavy-tailed data, maximum likelihood estimation, reliability analysis, hazard rate function, Rényi entropy, moment generating function, Newton-Raphson method, goodness-of-fit tests, Akaike information criterion, Bayesian information criterion, survival data modeling, likelihood ratio test
Cite Scienmag News
Reid Dalton. (November 20, 2026). New Transmuted Log-Logistic Models Improve Fitting for Complex Heavy-Tailed Data. Scienmag. https://scienmag.com/new-transmuted-log-logistic-models-improve-fitting-for-complex-heavy-tailed-data/
Reid Dalton. "New Transmuted Log-Logistic Models Improve Fitting for Complex Heavy-Tailed Data." Scienmag, 20 November 2026, https://scienmag.com/new-transmuted-log-logistic-models-improve-fitting-for-complex-heavy-tailed-data/. Accessed 20 November 2026.
Reid Dalton. "New Transmuted Log-Logistic Models Improve Fitting for Complex Heavy-Tailed Data." Scienmag. November 20, 2026. https://scienmag.com/new-transmuted-log-logistic-models-improve-fitting-for-complex-heavy-tailed-data/








