Evaluation of Karimi criterion in Markov processes

Salah H Abid

Nihad S Khala

Israa S Shihab

DOI: https://doi.org/10.47831/mjpas.v3i3.428


Abstract

One of the ways that can be exercised in autoregressive model order chosen is to select the order that
reduces the error of prediction. The final prediction error criterion employs this technique in order
selection. Regrettably, this criterion has poor performance in case of finite samples. Karimi 2007 derived a
criterion to address this problem. In this research, the Karimi criterion will be evaluated through the use of
some distributions. These distributions are Discrete Uniform, Cauchy, t and Log normal, in addition to
Gaussian distribution which is the basis of the Karimi criterion.