Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

EXPLORING THE JOINT USE OF PROPENSITY AND PROGNOSTIC SCORES FOR BIAS REDUCTION IN THE CONTEXT OF SMALL EDUCATIONAL PROGRAM EVALUATIONS

Tang, Yun (2018) EXPLORING THE JOINT USE OF PROPENSITY AND PROGNOSTIC SCORES FOR BIAS REDUCTION IN THE CONTEXT OF SMALL EDUCATIONAL PROGRAM EVALUATIONS. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img] PDF
Restricted to University of Pittsburgh users only until 26 June 2023.

Download (572kB) | Request a Copy

Abstract

Propensity and prognostic score methods are two statistical techniques used to correct for the selection bias in nonexperimental studies. Recently, the joint use of propensity and prognostic scores (i.e., two-score methods) has been proposed to improve the performance of adjustments using propensity or prognostic scores alone for bias reduction. The main purpose of this dissertation study was to evaluate the effectiveness of the joint use of propensity and prognostic scores for reducing bias of treatment effect estimates in quasi-experimental designs. To this end, a simulation study based on real educational data was conducted to investigate the comparative performance of separate and combined use of propensity and prognostic scores for recovering a simulated treatment effect under various conditions. These conditions were based on different control group sizes, outcome measures, and propensity score estimation methods. Specifically, four two-score methods were examined in this study: weighting and 1:1 optimal matching on the estimated prognostic propensity scores, and 1:1 and full matching on a Mahalanobis distance combining the estimated propensity and prognostic scores. Single score adjustments that were examined included 1:1 matching on the estimated propensity or prognostic scores, and weighting on the estimated propensity scores. The simulation results did not support the use of any of the two-score methods as alternatives to single score adjustments in estimation of treatment effects in the examined conditions. Instead, matching on the estimated prognostic scores showed some advantages over all the two-score methods and single score adjustments involving propensity scores only. However, this seemingly promising finding for adjustments on prognostic scores is tempered by the inherent “in-sample” problems for estimating prognostic scores.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Tang, Yunyut9@pitt.eduyut90000-0002-2714-9126
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Thesis AdvisorStone, Clementcas@pitt.educas
Committee MemberYe, Feifeifeifeiye@pitt.edufeifeiye
Committee MemberPage, Lindsaylpage@pitt.edulpage
Committee MemberTang, Fengyanfet7@pitt.edufet7
Date: 26 June 2018
Date Type: Publication
Defense Date: 27 February 2018
Approval Date: 26 June 2018
Submission Date: 26 June 2018
Access Restriction: 5 year -- Restrict access to University of Pittsburgh for a period of 5 years.
Number of Pages: 132
Institution: University of Pittsburgh
Schools and Programs: School of Education > Psychology in Education
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Bias reduction; Education; Observational data; Prognostic scores; Propensity scores; Quasi-experimental studies
Date Deposited: 26 Jun 2018 21:05
Last Modified: 26 Jun 2018 21:05
URI: http://d-scholarship.pitt.edu/id/eprint/34676

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item