Statistical contributions to non-experimental studies
Public DepositedAdd to collection
You do not have access to any existing collections. You may create a new collection.
Downloadable Content
Download PDFCitation
MLA
Choi, Byeongyeob. Statistical Contributions to Non-experimental Studies. Chapel Hill, NC: University of North Carolina at Chapel Hill Graduate School, 2015. https://doi.org/10.17615/9vfw-wd03APA
Choi, B. (2015). Statistical contributions to non-experimental studies. Chapel Hill, NC: University of North Carolina at Chapel Hill Graduate School. https://doi.org/10.17615/9vfw-wd03Chicago
Choi, Byeongyeob. 2015. Statistical Contributions to Non-Experimental Studies. Chapel Hill, NC: University of North Carolina at Chapel Hill Graduate School. https://doi.org/10.17615/9vfw-wd03- Last Modified
- March 19, 2019
- Creator
-
Choi, Byeongyeob
- Affiliation: Gillings School of Global Public Health, Department of Biostatistics
- Abstract
- The objective of this research is to develop the methods of statistical inference for a causal effect of an exposure on an outcome in the presence of unmeasured confounders. Instrumental variable (IV) analysis is frequently used to estimate exposure effect for the data with unmeasured confounding. For example, in a randomized clinical trial, subjects often fail to comply with their own treatment protocols and such a non-compliance may depend on unmeasured confounders. In this case, it is challenging to obtain the true treatment effect, which can be observed when all subjects comply their assigned regime. To obtain the true treatment effect, we may conduct IV analysis with a randomization indicator as an IV. In many randomized clinical trials or observational studies, incomplete outcomes such as survival times with censoring are obtained. There is a lack of IV methods for incomplete data such as survival data. Another tool to overcome unmeasured confounding is to use negative control outcomes. Negative control outcomes should satisfy specific conditions in casual relationships with the exposure, outcomes and confounders. Several studies have used negative control outcomes to determine the presence of unmeasured confounders. Especially, the approach to use negative control outcomes has been elegantly used by epidemiologists to identify unmeasured confounding in the studies of effectiveness of influenza vaccine on the elderly. However, statistical methods using negative control outcomes to obtain the estimator for causal effect of the exposure have not been investigated well. Another goal of this research is to develop improved confidence intervals for current status data. Confidence intervals for current status data have been well studied theoretically, their practical application has been limited, in part because of poor performance in small samples and in part because of computational difficulties. The subsampling-based method and likelihood-ratio test (LRT)-based method have been shown to have better coverage probabilities than a simple Wald-based method which may perform poorly in realistic sample sizes. However, those methods are complicated and require much more computational demands compared to Wald-based method. Therefore, we propose (1) Two-stage estimation of structural instrumental variable models with incomplete data, (2) Sensitivity analysis of regression results to unobserved confounding using a negative control outcome, (3) A new instrumental variable estimator using a negative control outcome and (4) Improved confidence intervals for current status data.
- Date of publication
- August 2015
- Keyword
- Subject
- DOI
- Identifier
- Resource type
- Rights statement
- In Copyright
- Advisor
- Ivanova, Anastasia
- Brookhart, Maurice
- Hudgens, Michael
- Fine, Jason
- Stürmer, Til
- Degree
- Doctor of Philosophy
- Degree granting institution
- University of North Carolina at Chapel Hill Graduate School
- Graduation year
- 2015
- Language
- Publisher
- Place of publication
- Chapel Hill, NC
- Access right
- There are no restrictions to this item.
- Date uploaded
- January 21, 2016
Relations
- Parents:
This work has no parents.
Items
Thumbnail | Title | Date Uploaded | Visibility | Actions |
---|---|---|---|---|
Choi_unc_0153D_15696.pdf | 2019-04-12 | Public | Download |