The Hjorth Model on Its Unit Support: Theory, Parameter Inference, Optimization, and Reliability Scenarios for Complex Data Modeling

Modeling bounded data presents a fundamental challenge in reliability and risk analysis, particularly when the underlying observations exhibit heterogeneous distributional shapes and complex failure-rate patterns. To address this challenge, we introduce a novel Unit Hjorth (UHj) distribution, obtained through an exponential transformation of the classical Hjorth model, which transfers its flexible reliability structure to the unit interval while retaining analytical tractability. A comprehensive theoretical investigation of the proposed model is developed, including its boundary behavior, limiting submodels, quantile function, ordinary moments, cumulants, mode characterization, order statistics, stress–strength reliability, and other important reliability measures. The UHj density can be strictly increasing or non-monotone, including unimodal forms, while its hazard rate can accommodate increasing, bathtub-shaped, upside-down-bathtub, and modified-bathtub patterns. This broad hazard-rate flexibility makes the model particularly suitable for representing heterogeneous reliability and risk profiles that cannot be adequately captured by conventional bounded distributions. For statistical inference, a comprehensive estimation framework is established based on maximum likelihood, maximum product of spacings, and six additional classical estimation methods. Their finite-sample performance is systematically assessed through extensive Monte Carlo simulations using multiple measures of bias, efficiency, accuracy, and numerical stability. The simulation results indicate that the likelihood- and spacing-based procedures generally provide the most accurate and stable parameter estimates. The practical relevance of the proposed model is further demonstrated through three real-world datasets from computer science, engineering, and environmental applications. In these applications, the UHj distribution provides competitive, and in several cases superior, goodness-of-fit performance compared with a range of established unit distributions. Overall, the proposed model provides a unified and flexible framework for bounded-data modeling, reliability assessment, risk characterization, and statistical inference, offering a useful addition to the class of bounded probability models for complex applied-data scenarios.