Contents Online
Statistics and Its Interface
Volume 11 (2018)
Number 2
Censored bimodal symmetric-asymmetric families
Pages: 237 – 249
DOI: https://dx.doi.org/10.4310/SII.2018.v11.n2.a3
Authors
Abstract
In this paper, we introduce two new families of distributions that are suitable for fitting unimodal as well as bimodal symmetric and asymmetric censored data. The models extend the skew normal model to bimodal symmetric and asymmetric situations and typically involves less parameters to be estimated than mixtures of normal distributions. Maximum likelihood estimation (MLE) is discussed and Fisher information matrices are derived. Results of a simulation study indicate stable parameter recovery in moderate and large samples. Applications to two real data sets are reported. The first data set is related to results of a study on antiretroviral therapy (HAART) to AIDS patients with strong evidence of bimodality and asymmetry. The second data set (fetal weight of unborn children) presents bimodal symmetry, well captured by the model introduced.
Keywords
bimodal distribution, generalized Gaussian distribution, kurtosis, power-normal model, skew-normal distribution, skewness
Received 25 October 2016
Published 7 March 2018