Abstract: The analysis of small datasets is a relevant problem in a lot of different fields, especially in medicine, and the use of appropriate resampling methods could provide better and more reliable results.
In this work, a new bagging survival tree model is proposed, where an extension of Efron’s bootstrap procedure is replaced in the classical model. The proper Bayesian bootstrap allows to enrich the original feature space with new observations sampled from a prior distribution, that are not already present in the original data.
Empirical results are shown through a sensitivity analysis and in a simulated study. The proposed model reaches competitive performances with respect to classical survival models (Cox model and survival random forest) in terms of integrated Brier score with higher stability. The biggest improvements are shown when small sample sizes are involved.
|
Keywords and phrases: survival analysis, bootstrap, Bayesian non-parametric learning, ensemble models.
Received: March 7, 2023; Accepted: April 22, 2023; Published: May 26, 2023
How to cite this article: Elena Ballante, An extension of generalized Bayesian ensemble tree models to survival analysis, Far East Journal of Theoretical Statistics 67(2) (2023), 137-146. http://dx.doi.org/10.17654/0972086323007
This Open Access Article is Licensed under Creative Commons Attribution 4.0 International License
References:
[1] M. Galvani, C. Bardelli, S. Figini and P. Muliere, A Bayesian nonparametric learning approach to ensemble models using the proper Bayesian bootstrap, Algorithms (Basel) 14(1) (2021), 11. [2] P. Muliere and P. Secchi, Bayesian nonparametric predictive inference and bootstrap techniques, Ann. Inst. Statist. Math. 48 (1996), 663-673. [3] B. Efron, Bootstrap methods: another look at the jackknife, Ann. Statist. 7(1) (1979), 1-26. [4] D. Rubin, The Bayesian bootstrap, Ann. Statist. 9(1) (1981), 130-134. [5] L. Gordon and R. Olshen, Tree-structured survival analysis, Cancer Treatment Reports 69(10) (1985), 1065-1069. [6] T. Hothorn, B. Lausen, A. Benner and M. Radespiel-Tröger, Bagging survival trees, Stat. Med. 23 (2004), 77-91. [7] H. Ishwaran, U. Kogalur, E. Blackstone and M. Laure, Random survival forest, The Annals of Applied Statistics 2(3) (2008), 841-860. [8] P. Andersen and R. Gill, Cox’s regression model for counting processes, a large sample study, Ann. Statist. 10 (1982), 1100-1120.
|