Bayesian Model Averaging for Propensity Score Analysis

Autor/inn/en	Kaplan, David; Chen, Jianshen
Institution	Society for Research on Educational Effectiveness (SREE)
Titel	Bayesian Model Averaging for Propensity Score Analysis
Quelle	(2013), (10 Seiten) PDF als Volltext kostenfreie Datei Verfügbarkeit
Sprache	englisch
Dokumenttyp	gedruckt; online; Monographie
Schlagwörter	Bayesian Statistics; Models; Probability; Monte Carlo Methods; Markov Processes; Longitudinal Studies; Surveys; Children; Kindergarten; Grade 1; Attendance; Early Childhood Longitudinal Survey + Suchen Sie Ihr Suchwort? Analogiemodell; Wahrscheinlichkeitsrechnung; Wahrscheinlichkeitstheorie; Monte-Carlo-Methode; Markowscher Prozess; Longitudinal study; Longitudinal method; Longitudinal methods; Längsschnittuntersuchung; Survey; Umfrage; Befragung; Child; Kind; Kinder; School year 01; 1. Schuljahr; Schuljahr 01; Anwesenheit
Abstract	The purpose of this study is to explore Bayesian model averaging in the propensity score context. Previous research on Bayesian propensity score analysis does not take into account model uncertainty. In this regard, an internally consistent Bayesian framework for model building and estimation must also account for model uncertainty. The significance of the current study is that it directly addresses the problem of uncertainty in propensity score models via the method of Bayesian model averaging (BMA). The usefulness of the proposed method is that it provides the investigator a way to incorporate prior knowledge regarding the relationship between the covariates and treatment selection (via the Kaplan and Chen, 2012 approach) while at the same time acknowledging model uncertainty via Bayesian model averaging. This paper provides a fully Bayesian MCMC methodology to obtain propensity score and treatment effect estimates, as well as R code to conduct such an analysis. Research design utilizes a combination of simulation studies and real data analysis. The simulation study examines the choice of parameter and model priors. The real data example examines a model relating full vs. part-day kindergarten attendance on achievement outcomes for first grade student using the ECLS-K. Preliminary findings suggest that the fully MCMC algorithm for Bayesian model averaging within the PSA framework provides accurate expected a posteriori estimates of the treatment effect. An appendix provides additional information on the two-step Bayesian PSA model, Bayesian model averaging, and computational considerations. (ERIC).
Anmerkungen	Society for Research on Educational Effectiveness. 2040 Sheridan Road, Evanston, IL 60208. Tel: 202-495-0920; Fax: 202-640-4401; e-mail: inquiries@sree.org; Web site: http://www.sree.org
Erfasst von	ERIC (Education Resources Information Center), Washington, DC
Update	2020/1/01