A flexible probability model for proportion data: Unit-half-normal distribution

Abstract

A new class of unimodal asymmetric distributions is introduced to the unit interval and these distributions are useful for modeling data of percentages, proportions and fractions. Therefore, we propose the unit-half-normal distribution as a contribution to the earlier path and investigate some of its mathematical properties. The maximum likelihood estimator is obtained with a comprehensive inference. This new class of distributions belongs to the exponential family, hence the uniformly minimum variance unbiased estimator of the distribution parameter is obtained. The distribution represents a power alternative to the unit interval distributions, namely the beta, Kumaraswamy and other recent ones. We investigate a small simulation study to analyze the behavior of the obtained estimators for different sample sizes. Moreover, we illustrate the goodness of fit of the proposed model for image data. Lastly, we describe a procedure of incorporating covariates into regression analysis of the proposed distribution.

Keywords:

• Half-normal distribution
• maximum likelihood estimation
• modeling
• minimum-variance unbiased estimator
• covariates into regression

Acknowledgments

The authors thank the anonymous reviewer for the constructive comments, which have contributed to improve the article.