Date on Master's Thesis/Doctoral Dissertation
5-2014
Document Type
Master's Thesis
Degree Name
M.S.
Department
Bioinformatics and Biostatistics
Committee Chair
Kong, Maiying
Subject
Statistical matching; Mathematical statistics
Abstract
Though randomized clinical (RCTs) trials are the gold standard for comparing treatments, they are often infeasible or exclude clinically important subjects, or generally represent an idealized medical setting rather than real practice. Observational data provide an opportunity to study practice-based evidence, but also present challenges for analysis. Traditional statistical methods which are suitable for RCTs may be inadequate for the observational studies. In this project, four of the most popular statistical methods for observational studies: ANCOVA, propensity score matching, regression with the propensity score as a covariate, and instrumental variables (IV) are investigated through application to MarketScan insurance claims data. Each of these methods is used to compare BMP versus autograft spinal surgeries for the outcomes length of stay, complications, and cost. Recommendations are made as to when each particular method may or may not be the optimal choice.
Recommended Citation
Gardner, Kristopher C. 1984-, "Statistical methods for assessing treatment effects for observational studies." (2014). Electronic Theses and Dissertations. Paper 479.
https://doi.org/10.18297/etd/479
