《PSM方法与应用》由会员分享,可在线阅读,更多相关《PSM方法与应用(21页珍藏版)》请在金锄头文库上搜索。
1、a105a105a105a105a105a105a105a105The Stata JournalEditorH. Joseph NewtonDepartment of StatisticsTexas A & M UniversityCollege Station, Texas 77843979-845-3142979-845-3144 FAXjnewtonstata-Executive EditorNicholas J. CoxDepartment of GeographyUniversity of DurhamSouth RoadDurham City DH1 3LEUnited King
2、domn.j.coxstata-Associate EditorsChristopher BaumBoston CollegeRino BelloccoKarolinska InstitutetDavid ClaytonCambridge Inst. for Medical ResearchCharles FranklinUniversity of Wisconsin, MadisonJoanne M. GarrettUniversity of North CarolinaAllan GregoryQueens UniversityJames HardinTexas A&M Universit
3、yStephen JenkinsUniversity of EssexJens LauritsenOdense University HospitalStanley LemeshowOhio State UniversityJ. Scott LongIndiana UniversityThomas LumleyUniversity of Washington, SeattleMarcello PaganoHarvard School of Public HealthSophia Rabe-HeskethInst. of Psychiatry, Kings College LondonJ. Pa
4、trick RoystonMRC Clinical Trials Unit, LondonPhilip RyanUniversity of AdelaideJeroen WeesieUtrecht UniversityJerey WooldridgeMichigan State UniversityCopyright Statement: The Stata Journal and the contents of the supporting files (programs, datasets, andhelp files) are copyright c by Stata Corporati
5、on. The contents of the supporting files (programs, datasets,and help files) may be copied or reproduced by any means whatsoever, in whole or in part, as long as anycopy or reproduction includes attribution to both (1) the author and (2) the Stata Journal.The articles appearing in the Stata Journal
6、may be copied or reproduced as printed copies, in whole or in part,as long as any copy or reproduction includes attribution to both (1) the author and (2) the Stata Journal.Written permission must be obtained from Stata Corporation if you wish to make electronic copies of theinsertions. This preclud
7、es placing electronic copies of the Stata Journal, in whole or in part, on publicallyaccessible web sites, fileservers, or other locations where the copy may be accessed by anyone other than thesubscriber.Users of any of the software, ideas, data, or other materials published in the Stata Journal or
8、 the supportingfiles understand that such use is made without warranty of any kind, by either the Stata Journal, the author,or Stata Corporation. In particular, there is no warranty of fitness of purpose or merchantability, nor forspecial, incidental, or consequential damages such as loss of profits
9、. The purpose of the Stata Journal is topromote free communication among Stata users.The Stata Technical Journal (ISSN 1536-867X) is a publication of Stata Press, and Stata is a registered trade-mark of Stata Corporation.a105a105a105a105a105a105a105a105The Stata Journal (2002)2,Number 4, pp. 358377E
10、stimation of average treatment eects basedon propensity scoresSascha O. BeckerUniversity of MunichAndrea IchinoEUIAbstract. In this paper, we give a short overview of some propensity scorematching estimators suggested in the evaluation literature, and we provide a setof Stata programs, which we illu
11、strate using the National Supported Work (NSW)demonstration widely known in labor economics.Keywords: st0026, propensity score, matching, average treatment eect, evaluation1IntroductionIn the evaluation literature, data often do not come from randomized trials but from(nonrandomized) observational s
12、tudies. In seminal work, Rosenbaum and Rubin (1983)proposed propensity score matching as a method to reduce the bias in the estimation oftreatment eects with observational datasets. These methods have become increasinglypopular in medical trials and in the evaluation of economic policy interventions
13、.Since in observational studies assignment of subjects to the treatment and controlgroups is not random, the estimation of the eect of treatment may be biased by theexistence of confounding factors. Propensity score matching is a way to “correct” theestimation of treatment eects controlling for the
14、existence of these confounding factorsbased on the idea that the bias is reduced when the comparison of outcomes is performedusing treated and control subjects who are as similar as possible. Since matching sub-jects on an n-dimensional vector of characteristics is typically unfeasible for large n,t
15、his method proposes to summarize pretreatment characteristics of each subject into asingle-index variable (the propensity score) that makes the matching feasible.In this paper, we give a short overview of some propensity score matching estimatorssuggested in the evaluation literature, and we provide
16、 a set of Stata programs, whichwe illustrate using the National Supported Work (NSW) demonstration widely knownin labor economics. In using these programs, it should be kept in mind that they onlyallow to reduce, and not to eliminate, the bias generated by unobservable confoundingfactors. The extent to which this bias is reduced depends crucially on the richnessand quality of the control variables on which the propensity score is computed and thematching pe