Title: Enriching the AR(p) process with skewed generalised normal innovations
Authors: JT Ferreira - University of Pretoria (South Africa) [presenting]
Ane Neethling - University of Pretoria (South Africa)
Andriette Bekker - University of Pretoria (South Africa)
Mehrdad Naderi - University of Pretoria (South Africa)
Abstract: The autoregressive (AR) process of order $p$ is of common value in many areas of research, not necessarily only in a time series environment. Usually, innovations are assumed to be iid normally distributed; this assumption may not characterise some true processes adequately due to the normal distributions restrictions regarding asymmetry. A skew generalised normal distribution is presented as an alternative to the usual normal assumption. It can account for not only skewness but also heavier tails that innovations might exhibit. This model is compared to previously proposed models, and its position as a valid contender as the choice of innovation process discussed via some simulation studies and data analysis.