Date on Master's Thesis/Doctoral Dissertation
Bioinformatics and Biostatistics
Biostatistics with a specialization in Decision Science, PhD
Committee Co-Chair (if applicable)
Mathematical statistics; Medicine--Research--Statistical methods; Materials science--Research
In this work we present three topics, each of which centered on either the application or modification of various linear regression methods. Our work with respect to the “Materials Genome” project while undermined by oversimplification and data integrity issues in its early stages, provides a sound platform from which the project can proceed successfully. Building upon a growing body of knowledge around the use of Weighted Generalized Estimating Equations (WGEE), our second investigation proposes an extension to that framework intended to address the inherent bias present in the analysis of clustered longitudinal data with potentially informative cluster sizes and temporal observation profiles. Having demonstrated the utility of our marginal WGEE’s with respect to mitigating induced bias our final investigation presents a comparison of our marginal WGEE’s to model estimation via Joint Likelihood maximization in certain simulation models. We find, as would be in line with expectation, comparable performance with a loss of efficiency in the marginal WGEE setting.
Bible, Joe, "Novel applications of and extensions to linear regression methods for the biomedical and materials sciences." (2015). Electronic Theses and Dissertations. Paper 2030.