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

Estimating probit models with self-selected treatments.

Jay Bhattacharya1, Dana Goldman, Daniel McCaffrey

  • 1Stanford Medical School, 117 Encina Commons, Stanford, CA 94305-6019, USA. jay@stanford.edu

Statistics in Medicine
|December 31, 2005
PubMed
Summary
This summary is machine-generated.

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 cost of dementia in the United States in 2026.

Alzheimer's & dementia : the journal of the Alzheimer's Association·2026
Same author

A dynamic microsimulation method for estimating dementia costs in the United States.

Alzheimer's & dementia : the journal of the Alzheimer's Association·2026
Same author

Insulin Costs and Use by Medicare Beneficiaries After the Inflation Reduction Act Out-of-Pocket Cap.

JAMA·2026
Same author

Evaluating evidence-based recruitment strategies for Alzheimer's disease and related dementias clinical trial research: A literature review.

The journal of prevention of Alzheimer's disease·2026
Same author

The new vision from the National Institute of Allergy and Infectious Diseases (NIAID).

Nature medicine·2026
Same author

Public Health.

Alzheimer's & dementia : the journal of the Alzheimer's Association·2025
Same journal

Predictor-Assisted Nonparametric Graphical Models With Multivariate Error-Prone Data.

Statistics in medicine·2026
Same journal

Optimizing Treatment Decision Estimation for Right-Censored Survival Data Through Parameter Transfer Learning.

Statistics in medicine·2026
Same journal

Latent Class Log-Linear Models for Estimating Diagnostic Test Accuracy Without a Gold Standard: A Simulation Study.

Statistics in medicine·2026
Same journal

Interpretable Bayesian Modeling for Multireader Multicase Studies: Addressing Overdispersion and Limited Sample Size in Diagnostic Enhancement Evaluation.

Statistics in medicine·2026
Same journal

Adaptive Sequential Multiple Hypotheses Testing for Concomitant Vaccine Safety Surveillance.

Statistics in medicine·2026
Same journal

Novel Distance Regression for Repeated Outcomes With Missing Data: Applications to Longitudinal and Crossover Studies of Microbiome Beta-Diversity.

Statistics in medicine·2026
See all related articles

Estimating treatment effects in non-randomized studies requires careful methods. Multivariate probit models are recommended over two-step procedures for binary outcomes, especially with complex data.

Area of Science:

  • Health outcomes research
  • Biostatistics
  • Econometrics

Background:

  • Non-randomized studies often involve self-selected samples, creating spurious correlations between treatment and outcome.
  • Standard binary dependent variable techniques (e.g., logit, probit) can yield inconsistent estimates in such settings.
  • Two-step procedures, analogous to two-stage least squares, are often considered but have limitations.

Purpose of the Study:

  • To evaluate the limitations of two-step procedures for estimating binary treatment effects on binary outcomes in non-randomized settings.
  • To compare the performance of two-step probit, two-stage least squares linear probability model, and multivariate probit estimators.
  • To provide guidance on selecting appropriate statistical methods for outcomes research.

Main Methods:

Related Experiment Videos

  • Monte Carlo simulations were conducted to compare estimator performance.
  • The study analyzed the consistency of different estimators under various data-generating processes.
  • An empirical example assessed the effect of insurance coverage on HIV+ patient mortality.

Main Results:

  • Two-step estimators were found to be generally inconsistent.
  • Multivariate probit estimators demonstrated superior performance compared to two-step and linear probability models.
  • The advantages of multivariate probit were particularly evident with multiple treatments, extreme outcome probabilities, or non-normal data.

Conclusions:

  • Multivariate probit models are recommended for estimating binary treatment effects on binary outcomes in non-randomized settings.
  • Standard two-step methods and linear probability models may produce biased results.
  • The choice of method significantly impacts the reliability of findings in outcomes research.