Publication:
A Bayesian model to estimate causality in PISA scores: a tutorial with application to ICT

Loading...
Thumbnail Image
Identifiers
Publication date
2015-07
Defense date
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Google Scholar
Export
Research Projects
Organizational Units
Journal Issue
Abstract
This paper presents a step-by-step tutorial to estimate causal effects in PISA 2012 by means of a nonparametric Bayesian modeling approach known as Bayesian Additive Regression Trees (BART), with an illustration of the causal impact of ICT on Spanish students' performance. The R code is explained in a way that can be easily applied to other similar studies. The application shows that, compared to more traditional methodologies, the BART approach is particularly useful when a high-dimensional set of confounding variables is considered as its results are not based on a sampling hypothesis. BART allows for the estimation of different interactive effects between the treatment variable and other covariates. BART models do not require the analyst to make explicit subjective decisions in which covariates must be included in the final models. This makes it an easy procedure to guide policy makers' decisions in different contexts
Description
Keywords
Causality in education, BART models, Propensity score
Bibliographic citation