TY - GEN
T1 - On model selection by quasi-Bayesian predictive densities
AU - Djurić, Petar M.
AU - Doroslovački, Miloš
N1 - Publisher Copyright:
© 1992 IEEE.
PY - 1992
Y1 - 1992
N2 - A Bayesian procedure for model selection recently proposed by Aitkin [1] drew the attention by its heretical (from a Bayesian point of view) feature to validate a model by data that have already been used to estimate the model parameters. In the paper we discuss some of the issues that this procedure was meant to resolve. We also show how the same issues could be resolved without violating the Bayesian paradigm. Examples and computer simulations are provided in which the Bayesian and quasi-Bayesian approaches are compared. It is shown that in the case of linear models the quasi- Bayesian methods yield poor results.
AB - A Bayesian procedure for model selection recently proposed by Aitkin [1] drew the attention by its heretical (from a Bayesian point of view) feature to validate a model by data that have already been used to estimate the model parameters. In the paper we discuss some of the issues that this procedure was meant to resolve. We also show how the same issues could be resolved without violating the Bayesian paradigm. Examples and computer simulations are provided in which the Bayesian and quasi-Bayesian approaches are compared. It is shown that in the case of linear models the quasi- Bayesian methods yield poor results.
UR - https://www.scopus.com/pages/publications/85067259266
U2 - 10.1109/ISCAS.1992.230541
DO - 10.1109/ISCAS.1992.230541
M3 - Conference contribution
AN - SCOPUS:85067259266
T3 - Proceedings - IEEE International Symposium on Circuits and Systems
SP - 2372
EP - 2375
BT - 1992 IEEE International Symposium on Circuits and Systems, ISCAS 1992
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 1992 IEEE International Symposium on Circuits and Systems, ISCAS 1992
Y2 - 10 May 1992 through 13 May 1992
ER -