Answers are grounded in this paper and cite their sources.
This study developed a Modified Bivariate Conditional Lomax Distribution using Power-Link Function (MBCLD-PLF) in which the scale parameter is linked to the conditioning variable through a power function. The model was proposed to provide a simple but flexible way of describing dependence between two positive heavy-tailed variables. Instead of introducing dependence mainly through a copula, the proposed model allows the conditional scale parameter to change directly with the value of the first variable. A Monte Carlo simulation with 1,000 replications at each sample size of 50, 100, 150, 300, 500 and 1,000 was used to examine the finite-sample behaviour of the maximum likelihood estimators. The results showed that estimation generally improved as the sample size increased, with lower bias, mean squared error and root mean squared error, while the confidence-interval coverage probabilities remained close to the nominal 95% level. Numerical optimisation converged successfully across the simulation settings, although boundary estimates were more frequent at the smallest sample size. The model was also fitted to the Loss–ALAE insurance dataset containing 1,500 paired observations. Its performance was compared with the Classical Bivariate Lomax, Bivariate Power Lomax–FGM, Clayton Copula Bivariate Lomax and Bivariate Power Lomax–Sarmanov models. The proposed model recorded the highest maximised log-likelihood and the lowest AIC, BIC, CAIC and HQIC. However, the Clayton and Sarmanov models produced a smaller Kolmogorov–Smirnov, Anderson–Darling and Cramér–von Mises Rosenblatt-residual statistics. The findings therefore support MBCLD-PLF as a competitive conditional alternative with strong likelihood-based performance for heavy tailed data.
No takes yet. Share an insight, caveat, or question.
Audu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: