Functional Regression Methods for Densely-Sampled Biomarkers in the ICU

Intensive care unit;functional data analysis;longitudinal data analysis;survival analysis;biomarkers;acute respiratory distress syndrome;variable-domain functional regression;historical Cox model;penalized Cox regression;pcox;refund;Biostatistics

Gellar, Jonathan EvanGuallar, Eliseo ;

Johns Hopkins University

关键词: Intensive care unit;    functional data analysis;    longitudinal data analysis;    survival analysis;    biomarkers;    acute respiratory distress syndrome;    variable-domain functional regression;    historical Cox model;    penalized Cox regression;    pcox;    refund;    Biostatistics;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/39384/GELLAR-DISSERTATION-2015.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

This thesis develops methods for modeling longitudinal predictors by treating them as functional covariates in regression models. First, I introduce Variable-Domain Functional Regression, which extends the generalized functional linear model by allowing for functional covariates that have subject-specific domain widths. I then propose a blueprint for the inclusion of baseline functional predictors in Cox proportional hazards models. Finally, I propose the Historical Cox Model, which introduces a new way of modeling time-varying covariates in survival models by including them as historical functional terms. Methods were motivated by and applied to a study of association between daily measures of the Intensive Care Unit (ICU) Sequential Organ Failure Assessment (SOFA) score and mortality, and are generally applicable to a large number of new studies that record a continuous variables over time.