Optimal designs for observational studies using integer programming.

Format:

Book

Thesis/Dissertation

Author/Creator:

Zubizarreta, José R.

Contributor:

Silber, Jeffrey H., committee member.

Small, Dylan S., committee member.

Rosenbaum, Paul R., advisor.

University of Pennsylvania. Statistics.

Language:

English

Subjects (All):

Statistics.

0463.

Penn dissertations--Statistics.

Statistics--Penn dissertations.

Local Subjects:

Penn dissertations--Statistics.

Statistics--Penn dissertations.

0463.

Physical Description:

141 pages

Contained In:

Dissertation Abstracts International 75-01B(E).

System Details:

Mode of access: World Wide Web.

text file

Summary:

This thesis is organized around four papers that present and illustrate new methods for optimal designs in observational studies. The first paper proposes a new matching method based on mixed integer programming that directly targets covariate balance. The second paper describes a new related method based on integer programming that can be used to strengthen an instrumental variable while adjusting very precisely for observed covariates. The third paper introduces cardinality matching and shows how this method can be used to reduce heterogeneity and thus sensitivity to unobserved bias. The fourth and final paper illustrates the use of matching based on mixed integer programming and the theory of design sensitivity to estimate the effect of the 2010 Chilean earthquake on posttraumatic stress.

Notes:

Thesis (Ph.D. in Statistics) -- University of Pennsylvania, 2013.

Source: Dissertation Abstracts International, Volume: 75-01(E), Section: B.

Adviser: Paul R. Rosenbaum.

Local Notes:

School code: 0175.

ISBN:

9781303397066

Access Restriction:

Restricted for use by site license.