ARCH Model Evaluation and Selection
Evdokia Xekalaki, Stavros Antonios Degiannakis · 2010
This chapter gives a concise presentation of aspects of ARCH model evaluation and selection among a set of competitors. At the same time, it provides a series of handy programs for exploring the various theoretical aspects of model evaluation and selection. The evaluation of models is viewed in terms of information criteria, and statistical loss functions. Loss functions that are dependent upon the aims of a specific usage/application are discussed in the chapter. A series of EViews programs that simulate data from an ARCH process, estimate several ARCH models, including the one used to generate the data, and then assess the ability of some of the mentioned model evaluation methods in ‘picking’ the true data-generating process is presented. The chapter also investigates use of the mean or the median value of the loss functions. Controlled Vocabulary Terms Akaike information criterion; autoregressive conditional heteroskedasticity