Sensitivity analysis in observational studies / Liansheng Wang.

Format:

Book

Manuscript

Microformat

Thesis/Dissertation

Author/Creator:

Wang, Liansheng.

Contributor:

Krieger, Abba M., advisor.

University of Pennsylvania.

Language:

English

Subjects (All):

Penn dissertations--Statistics.

Statistics--Penn dissertations.

Penn dissertations--Managerial science and applied economics.

Managerial science and applied economics--Penn dissertations.

Local Subjects:

Penn dissertations--Statistics.

Statistics--Penn dissertations.

Penn dissertations--Managerial science and applied economics.

Managerial science and applied economics--Penn dissertations.

Physical Description:

vii, 100 pages ; 29 cm

Production:

2001.

Summary:

This thesis considers observational studies in which experimental units are not randomly assigned to treatment and control. Two sources of bias in drawing inferences may exist. One source of bias is that treated subjects might differ from control subjects in terms of observed covariates. A common way of dealing with this bias is by matching treated subjects to controls that are similar in terms of these covariates. Another source of bias, often referred to as hidden bias, occurs when treated subjects differ from control subjects in variables that are unobserved. Various models that relate the unobserved covariates to treatment assignment and response have been proposed in the literature. These models provide for a "sensitivity analysis", an assessment of how sensitive are the inferences that are made to the potential existence of hidden bias. This thesis has two primary objectives. The models that relate observed values to potentially unobserved covariates are extended. In addition, sensitivity analysis results are expanded to additional situations of practical importance. All of these methods are illustrated with real-world examples. The thesis concludes with further extensions along the lines already considered in this dissertation.

Notes:

Supervisor: Abba M. Krieger.

Thesis (Ph.D. in Statistics) -- University of Pennsylvania, 2001.

Includes bibliographical references.

Local Notes:

University Microfilms order no.: 3003705.

OCLC:

244972411