B1252
Title: Parameter estimation of the partially linear models with skew heavy-tailed error distributions
Authors: Fatma Zehra Dogru - Giresun University (Turkey)
Olcay Arslan - Ankara University (Turkey) [presenting]
Abstract: Partially linear models are considered an important flexible generalization of the linear model in applications for modelling economic and biometric datasets. Partially linear models contain a non-parametric component of some covariate besides the linear parametric part of the model. In general, the error term in a partially linear model is assumed to have a normal distribution. However, in applications, datasets may not have a normal distribution, so modelling a partially linear model under the assumption of normality may not be appropriate. Partially linear models under the skew-normal distribution were proposed to deal with the skewness in the data sets. However, the dataset may also have a heavy-tailedness problem along with the skewness. Therefore, we will propose modelling partial linear models with the skewed Laplace normal distribution. The skew Laplace normal distribution is a heavy-tailed alternative to the skew-normal distribution with the same number of parameters. We provide an expectation-maximization (EM) algorithm for the maximum likelihood estimation procedure of the proposed skew Laplace Normal partially linear models. We conduct a simulation study and a real data example to demonstrate the performance of the proposed model.