An analysis on covariates selection problem for Gaussian model by Maximum a posteriori criterion using frequentist and Bayesian approaches

Document Type : Original Paper

Authors

1 Department of Statistics, Faculty of Mathematic and Computer Science, Amirkabir University of Technolgy, Tehran, Iran

2 Department of Statistics, Faculty of Mathematical Sciences, Shahid Beheshti University, Tehran, Iran

Abstract

Choosing the most suitable fitted model on data is one of the common challenges in statistical modeling. Maximum a posteriori (MAP) criterion is a method used in both frequentist and Bayesian approaches. Additionally, the utility of the model is used as a tool to compare the performances of methods. In this paper, the MAP method is applied for the Gaussian model and its performance is compared to that of frequentist approach. Also, an analytical form of utility estimation is proposed. Besides, using simulation studies, it is shown that the Gaussian model has better performance, based on both utility and mean of squared errors (MSE) criteria, when it is used by the Bayesian approach. However, both frequentist and Bayesian approaches avoid over-fitting by increasing the sample size. Also, by increasing correlation among covariates, MSE increases, while the tendency of choosing fewer covariates is raised. Eventually, the study on a real dataset is shown that in both frequentist and Bayesian approaches, MSE of selected models decreases when the size of sample increases.

Keywords

Main Subjects

References

1-Geisser, S. and Eddy, W.F. (1979). A predictive approach to model
selection, Journal of the American Statistical Association, 74(365), 153–
160.
2- Watanabe, S. (2009). Algebraic geometry and statistical learning theory
(Vol. 25), Cambridge University Press.
3- Spiegelhalter, D. J., Best, N. G., Carlin, B. P., and Van Der Linde, A.
(2002). Bayesian measures of model complexity and fit. Journal of the
Royal Statistical Society: Series b (Statistical Methodology), 64(4), 583-
639.
4- Laud, P.W. and Ibrahim, J.G. (1995). Predictive model selection, Journal
of the Royal Statistical Society. Series B (Methodological), 57(1), 247–
262.
5-Gelfand, A.E. and Ghosh, S.K. (1998). Model choice: a minimum posterior
predictive loss approach, Biometrika, 85(1), 1–11. 
Journal of Advanced Mathematical Modeling
Volume 10, Issue 2
December 2020
Pages 245-266

Files

History

  • Receive Date: 26 August 2019
  • Revise Date: 25 February 2020
  • Accept Date: 12 March 2020

Share

How to cite

Statistics