Application of the Topp-Leone Kumaraswamy Half-Logistic Distribution
Eleta Oluseyi Kelvin *
Department of Statistics, Faculty of Physical Sciences, Federal University, Lokoja, Nigeria.
Eunice Ohunene Job
Department of Statistics, Faculty of Physical Sciences, Federal University, Lokoja, Nigeria.
Benson Onoghojobi
Department of Statistics, Faculty of Physical Sciences, Federal University, Lokoja, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
This study develops and examines the four-parameter Topp-Leone Kumaraswamy Half-Logistic (TLKHL) distribution, obtained by applying the Topp-Leone Kumaraswamy-G generator to the classical Half-Logistic distribution. The model is intended to provide additional flexibility for analysing positive continuous observations with different distributional and hazard-rate characteristics. Its cumulative distribution function and probability density function are specified, and mathematical properties, including series representations, probability-weighted moments, ordinary moments, a moment-generating function, reliability and hazard-rate functions, and a quantile function, are investigated. The model parameters are estimated by maximum likelihood, with nonlinear estimating equations solved using numerical optimisation. A Monte Carlo study uses 1,000 replications at sample sizes of 20, 50, 100, 250, 500 and 1,000 for two selected parameter configurations. The reported results generally show smaller empirical biases and root mean squared errors at larger sample sizes. The model is also applied to 69 measurements of carbon-fibre tensile strength, expressed in gigapascals, obtained using a gauge length of 20 mm. Comparisons with seven alternative distributions are based on information criteria and goodness-of-fit statistics. In the reported comparison, the TLKHL model has the lowest Akaike Information Criterion, whereas the lognormal model has the lowest Bayesian Information Criterion and reported Kolmogorov–Smirnov statistic. These findings suggest that the TLKHL distribution is a competitive candidate for modelling the positive continuous data examined, although model preference depends on the assessment criterion.
Keywords: Topp-Leone Kumaraswamy-G family, half-logistic distribution, lifetime distribution, maximum likelihood estimation, monte carlo simulation, reliability analysis, goodness-of-fit