A0815
Title: Robust model selection in mixture regression
Authors: Sollie Millard - University of Pretoria (South Africa) [presenting]
Salomi Millard - University of Pretoria (South Africa)
Frans Kanfer - University of Pretoria (South Africa)
Andre Kleynhans - University of Pretoria (South Africa)
Abstract: A penalised likelihood approach combined with self-paced learning is proposed for model selection in mixtures of regression models. This approach mitigates the impact of outliers. An EM-type algorithm is proposed in order to maximize the penalised log-likelihood function. The properties of the estimation algorithm are demonstrated using a simulation approach. The proposed estimation approach is also illustrated using real data.
