Constructing Propensity score weighted and matched Samples Stacey L

Slides:



Advertisements
Similar presentations
REGRESSION, IV, MATCHING Treatment effect Boualem RABTA Center for World Food Studies (SOW-VU) Vrije Universiteit - Amsterdam.
Advertisements

1 Arlene Ash QMC - Third Tuesday September 21, 2010 (as amended, Sept 23) Analyzing Observational Data: Focus on Propensity Scores.
Propensity Score Matching Lava Timsina Kristina Rabarison CPH Doctoral Seminar Fall 2012.
Impact Evaluation: The case of Bogotá’s concession schools Felipe Barrera-Osorio World Bank 1 October 2010.
EVIDENCE BASED MEDICINE
Matching Methods. Matching: Overview  The ideal comparison group is selected such that matches the treatment group using either a comprehensive baseline.
Propensity Score Matching and Variations on the Balancing Test Wang-Sheng Lee Melbourne Institute of Applied Economic and Social Research The University.
Article Review Cara Carty 09-Mar-06. “Confounding by indication in non-experimental evaluation of vaccine effectiveness: the example of prevention of.
Types of Research and Designs This week and next week… Covering –Research classifications –Variables –Steps in Experimental Research –Validity –Research.
Matching Estimators Methods of Economic Investigation Lecture 11.
Chapter 7 Sampling Distributions Statistics for Business (Env) 1.
AFRICA IMPACT EVALUATION INITIATIVE, AFTRL Africa Program for Education Impact Evaluation David Evans Impact Evaluation Cluster, AFTRL Slides by Paul J.
Generalizing Observational Study Results Applying Propensity Score Methods to Complex Surveys Megan Schuler Eva DuGoff Elizabeth Stuart National Conference.
Can Mental Health Services Reduce Juvenile Justice Involvement? Non-Experimental Evidence E. Michael Foster School of Public Health, University of North.
Matching STA 320 Design and Analysis of Causal Studies Dr. Kari Lock Morgan and Dr. Fan Li Department of Statistical Science Duke University.
Using Propensity Score Matching in Observational Services Research Neal Wallace, Ph.D. Portland State University February
Chapter Eight: Quantitative Methods
Considering model structure of covariates to estimate propensity scores Qiu Wang.
MATCHING Eva Hromádková, Applied Econometrics JEM007, IES Lecture 4.
(ARM 2004) 1 INNOVATIVE STATISTICAL APPROACHES IN HSR: BAYESIAN, MULTIPLE INFORMANTS, & PROPENSITY SCORES Thomas R. Belin, UCLA.
Patricia Gonzalez, OSEP June 14, The purpose of annual performance reporting is to demonstrate that IDEA funds are being used to improve or benefit.
Impact Evaluation Methods Regression Discontinuity Design and Difference in Differences Slides by Paul J. Gertler & Sebastian Martinez.
Research and Evaluation Methodology Program College of Education A comparison of methods for imputation of missing covariate data prior to propensity score.
Looking for statistical twins
Eastern Michigan University
Chapter 11: The Nuts and Bolts of one-factor experiments.
Sample Size Determination
An introduction to Survival analysis and Applications to Predicting Recidivism Rebecca S. Frazier, PhD JBS International.
CHAPTER 4 Designing Studies
Ling Ning & Mayte Frias Senior Research Associates Neil Huefner
EXPERIMENTAL RESEARCH
Measuring the Effects of an Irrigation and Land Tenure Security Initiative in the Senegal River Valley Baseline findings and evaluation challenges March.
Patient Registries and Health Outcomes in Diabetes: A Retrospective Study Nipa Shah, MD1; Fern Webb, PhD1; Liane Hannah, BSH1; Carmen Smotherman, MS2;
Lecture Slides Elementary Statistics Twelfth Edition
Statistical Core Didactic
Sec 9C – Logistic Regression and Propensity scores
Experimental Research Designs
Experimental Design-Chapter 8
Coding Manual and Process
Between-Subjects, within-subjects, and factorial Experimental Designs
An Evaluation of Rhode Island’s CTC PCMH Program
TITLE IV-E WAIVER SITE VISIT
Chapter 13- Experiments and Observational Studies
March 2017 Susan Edwards, RTI International
Analyzing Intervention Studies
Chapter Eight: Quantitative Methods
S1316 analysis details Garnet Anderson Katie Arnold
Date: Presenter: Ryan Chen
Impact evaluation: The quantitative methods with applications
Matching Methods & Propensity Scores
Propensity Score Matching Makes Program Evaluation Easy
Matching Methods & Propensity Scores
Methods of Economic Investigation Lecture 12
Cross Sectional Designs
Treatment effect: Part 2
How Psychologists Do Research
Conducting Propensity Score Matching and Survival Analysis to Predict Recidivism for a Home Visitation Program Case Study, and Applying Results Propensity.
Impact Evaluation Methods
SPSS Propensity Score Matching: An overview
Matching Methods & Propensity Scores
Introduction to Experimental and Observational Study Design
RCT Workshop, Bushehr 2/16/2019.
Impact Evaluation Methods: Difference in difference & Matching
Evaluating Impacts: An Overview of Quantitative Methods
Experiments II: Validity and Design Considerations
The European Statistical Training Programme (ESTP)
Chapter: 9: Propensity scores
Improving Overlap Farrokh Alemi, Ph.D.
Sample Sizes for IE Power Calculations.
Reminder for next week CUELT Conference.
Presentation transcript:

Constructing Propensity score weighted and matched Samples Stacey L Constructing Propensity score weighted and matched Samples Stacey L. Houston, II, MA JBS International Propensity score matching

Introduction to Use of Propensity Scores (P Scores) Non-experimental designs and observational studies lack random assignment Selection bias: large differences in group characteristics Influence estimated group effects Methods for adjusting include: Propensity Score Weighting (PSW) and Propensity Score Matching (PSM) Specify that bias based on who selects into treatment – there might be characteristics of individuals in the treatment group that make them more likely to go into treatment – for example people who are more econ stable may be more likely to enter treatment because they have more time and resources to enter treatment which could influence the outcome

An Example Home Visitation No Home Visitation

What is a Propensity Score? An estimate of the probability of finding the subject in a particular group of interest Logistic regression can predict likelihood of being in a particular group given individual covariates Covariates to include Variables hypothesized to be associated with both treatment and outcome All available data Typically use longitudinal data with baseline characteristics in match

An Example Home Visitation No Home Visitation .8 .8 .2 .2 .2 .8 .8 .8 Important to emphasize .8 .8 Home Visitation No Home Visitation

Propensity Score Uses Propensity Score Weighting (PSW) Pros: retain entire sample Cons: difficult explanation Propensity Score Matching (PSM) Pros: simple representation of data; lower variance in treatment effect estimate; can use any analytic method Cons: sample size dependent; requires sufficient overlap between groups Any analytic method: Meaning once you have the matched sample you can then conduct any analysis that youre interested in such as regression, survival analysis, Hierarchical modeling Match can occur whether it’s a sufficient match or not

Weighting - Unweighted .8 .8 .8 .8 .2 .2 .8 .2 .8 .8 .8 .8 Home Visitation No Home Visitation

Weighting - Weighted Home Visitation No Home Visitation IPW=1/1-ps 1.25 5 5 1.25 1 5 5 1.25 1.25 Make clear that first slide is p score and p score and emphasizing that weight is create differently for both groups 1.25 1.25 5 5 1 Home Visitation No Home Visitation

Matching - Unmatched Home Visitation No Home Visitation .8 .8 .8 .8 .2 .8 .8 .2 .2 .8 .2 .8 .8 .8 .8 Home Visitation No Home Visitation

Matching – Matched (Exact) .8 .8 .8 .8 .2 .2 .2 .8 .8 .8 .8 .8 Home Visitation No Home Visitation

Complexities Multiple Covariates/Characteristics Matching Methods Exact matching Nearest neighbor Caliper matching With(out) replacement Multiple Treatment Groups Another key issue is whether controls can be used as matches for more than one treated unit; whether the matching should be done “with replacement” or “without replacement.” Matching with replacement can often yield better matches because controls that look similar to many treated units can be used multiple times. Additionally, like optimal matching, when matching with replacement the order in which the treated units are matched does not matter. However, a drawback of matching with replacement is that it may be that only a few unique control units will be selected as matches; the number of times each control is matched should be monitored Do nearest neighbor example with or without replacement

An Example Are families who participate in B&B less likely to have a subsequent child protective services referral than those who do not?

Propensity Score Weighting Table 1. Means, Percentages, and Standard Deviations (SD) for All Study Variables Unweighted   Propensity Score Weighted Treatment Control (N=893) (N=9,708) Mean/ Variables Percent Demographics Person was White (versus nonwhite) 28.20 36.80 *** 37.40 36.10 Person's primary language was English (versus any other) 82.80 82.90 81.40 Age in 2015 27.40 30.54 30.94 30.29 ** Unique Referrals At/Before Point of Eligibility for the Program Number of Unique Perpetrator or Other (restricted to those when person was 17 or older) Referrals 3.10 2.04 2.41 2.14 Number of Unique Victim Referrals 2.45 0.66 0.81 0.82 Age of Person's First Referral 16.71 23.76 23.61 23.17 *p<.05; **p<.01; ***p<.001 (treatment compared to control group). Maybe just do demographics! Probably just do the two means Mention the sample sizes and how we get to use the full sample Take some time to orient them to the table itself – put a white block over the weighted

Propensity Score Matching Table 1. Means, Percentages, and Standard Deviations (SD) for All Study Variables Unmatched Matched 2:1 Treatment Control (N=496) (N=9,210) (N=493) (N=985) Mean/   Variables Percent Demographics Person was White (versus nonwhite) 29.20 36.70 *** 29.40 30.60 Person's primary language was English (versus any other) 83.70 82.90 83.60 87.00 Age in 2015 27.46 30.39 27.48 27.54 Unique Referrals At/Before Point of Eligibility for the Program Number of Unique Perpetrator or Other (restricted to those when person was 17 or older) Referrals 2.94 1.92 2.87 2.61 Number of Unique Victim Referrals 2.41 0.70 2.35 Age of Person's First Referral 16.56 23.67 16.63 16.62 *p<.05; **p<.01; ***p<.001 (treatment compared to control group). Maybe just do demographics! Probably just do the two means

Propensity Score Best Practices If covariates are known to impact outcome variable, be sure to include them when possible Always demonstrate level of imbalance both before and after matching or weighting Try to use P scores only when imbalance is high initially Iteratively check balance and respecify propensity score regression requires sufficient overlap between groups

Summary Propensity scores can be an effective method of reducing selection bias DO NOT match or weight just because of recent trends Even if p-scores are not used, still a beneficial practice to estimate p-scores for group assignment of interest Especially circumstances when random assignment isn’t feasible and you have sufficiently available comparison group data and when you follow

Thank you! Questions? Additional resources and step by step instructions for conducting a survival analysis can be found at: