Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Propensity score estimation with boosted regression for evaluating causal effects in observational studies.

Daniel F McCaffrey1, Greg Ridgeway, Andrew R Morral

  • 1Public Safety and Justice Program, Drug Policy Research Center, RAND Corporation, 201 North Craig Street, Pittsburgh, PA 15213, USA. danielm@rand.org

Psychological Methods
|December 16, 2004
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

The Science of Gun Policy: A Critical Synthesis of Research Evidence on the Effects of Gun Policies in the United States, Fifth Edition.

Rand health quarterly·2026
Same author

Disparities in the prevalence and reporting of civilian justifiable firearm homicide.

Injury epidemiology·2026
Same author

Racial and ethnic disparities in fatal police shootings: Variation across U.S. states and the role of firearm ownership.

PloS one·2026
Same author

State-Level Estimates of Nonfatal Firearm Injury Hospitalizations, 2000-2021.

American journal of public health·2025
Same author

Practical challenges in mediation analysis: a guide for applied researchers.

Health services & outcomes research methodology·2025
Same author

The Science of Gun Policy: A Critical Synthesis of Research Evidence on the Effects of Gun Policies in the United States, Fourth Edition.

Rand health quarterly·2024

Boosting improves causal effect modeling using observational data by accurately estimating propensity scores. This method reduces pretreatment differences, revealing true treatment effects in studies like adolescent substance abuse programs.

Area of Science:

  • Causal inference
  • Statistical modeling
  • Observational data analysis

Background:

  • Causal effect modeling with naturalistic data is difficult due to confounding pretreatment characteristics.
  • Propensity score methods can address confounds but face challenges with numerous covariates and complex associations.

Purpose of the Study:

  • To demonstrate how boosting, a statistical technique, can overcome challenges in propensity score estimation.
  • To apply boosting to observational data for more accurate causal effect modeling.

Main Methods:

  • Utilized boosting, a modern statistical technique, for propensity score estimation.
  • Applied the method to a study of adolescent probationers in substance abuse treatment programs.
  • Estimated propensity score weights using boosting.

Related Experiment Videos

Main Results:

  • Boosting effectively eliminated most pretreatment group differences.
  • The estimated propensity score weights substantially altered the apparent relative effects of adolescent substance abuse treatment.
  • Demonstrated the utility of boosting in handling complex covariate structures and uncertain functional forms.

Conclusions:

  • Boosting offers a robust solution for accurate propensity score estimation in observational studies.
  • This approach enhances causal inference by mitigating confounding bias.
  • The findings have significant implications for understanding treatment effectiveness in real-world settings.