A0176
Title: Inference and diagnostics for a heteroscedastic partially linear model with skew heavy-tailed error distribution
Authors: Fatma zehra Dogru - Giresun University (Turkey) [presenting]
Olcay Arslan - Ankara University (Turkey)
Abstract: Partially Linear Models (PLMs) have gained attention from researchers as valuable tools for dealing with diverse data sets, particularly in fields such as economics and biometrics. Traditionally, PLMs assume a normal distribution for the error term. However, real-world data often deviate from this assumption, exhibiting skewness, heavy-tailed, and heteroscedasticity. Therefore, the Heteroscedastic Partial Linear Model (HPLM) under the Skew Laplace Normal (SLN) distribution is investigated. This model aims to address skewness, heavy-tailing, and heteroscedasticity simultaneously. The model parameters are estimated by Maximum Likelihood Estimation (MLE) using the Expectation/Conditional Maximisation (ECM) algorithm. The influence of diagnostics tailored to the HPLM-SLN model is also examined. In addition, a likelihood ratio test is introduced to assess the homogeneity of the scale parameter, providing insight into the variance homogeneity assumptions. Extensive simulation studies are performed to evaluate the performance of the ECM algorithm and the likelihood ratio test in terms of variance homogeneity. Finally, the effectiveness of the HPLM-SLN model is demonstrated through its application to a real-world dataset on ragweed pollen concentration, demonstrating its utility in practical scenarios.
