|
|
 |
| |
|
|
|
A Novel Unit Modified Weibull Distribution: Properties, Quantile Regression and Applications |
|
|
|
PP: 1559-1594 |
|
|
doi:10.18576/jsap/150605
|
|
|
|
Author(s) |
|
|
|
Abdul Ghaniyyu Abubakari,
|
|
|
|
Abstract |
|
|
| In this paper, we introduce a novel unit modified Weibull distribution for modelling data defined on the interval (0,1). Such data are recorded across disciplines ranging from engineering to the social sciences, and often exhibit skewness, outliers and non- monotone hazard rates that standard unit distributions cannot adequately capture. This has motivated the continuous development of new unit distributions for modelling complex data on the unit interval. The density of the proposed distribution can be increasing, decreasing, or unimodal with left or right skewness, while its hazard rate can exhibit monotone increasing, monotone decreasing, bathtub and modified bathtub shapes. Some properties, including quantile function, moments, stochastic ordering, order statistics and inequality measures, are derived. Also, the identifiability of its parameters is established. Maximum likelihood estimators are obtained for the parameters, and simulations reveal that the estimators are consistent and asymptotically unbiased. An alternative quantile based regression model using the proposed distribution is developed. Four real data applications demonstrate that the distribution and its regression model provide better fit than nine established distributions and four quantile regression models, respectively. The proposed methodology expands the toolkit available for unit data modeling.
|
|
|
|
|
 |
|
|