Home
Resource
Power analysis in longitudinal one-way crossover studies
PDF
PDF format is widely accepted and good for printing.
Plug-in required
PDF-1(1.46 MB)
Citation & Export
View Usage Statistics
Staff View

Citation & Export
Hide

Simple citation

Hu, Yirui. Power analysis in longitudinal one-way crossover studies. Retrieved from https://doi.org/doi:10.7282/T3Q52RX9

Export

Click here for information about Citation Management Tools at Rutgers.

Statistics
Hide

Description

TitlePower analysis in longitudinal one-way crossover studies
NameHu, Yirui (author); Hoover, Donald R (chair); Hung, Ying (internal member); Xiao, Han (internal member); Shi, Qiuhu (outside member); Rutgers University; Graduate School - New Brunswick
Date Created2016
Other Date2016-10 (degree)
SubjectStatistics and Biostatistics
Extent1 online resource (x, 131 p. : ill.)
DescriptionThis dissertation develops a general power estimation framework to estimate the variance of the new intervention effect estimate for longitudinal one-way crossover designs. Orthogonalized decomposition is developed for compound symmetry correlation of repeated measurements over time. In particular, we merge conventional difference-in-differences (DD) and more newly developed general stepped-wedge (SW) studies for both randomized and non-randomized allocation of units to the intervention, and investigate on the optimality properties in terms of study power (i.e. minimum variance of the intervention effect estimate). For a fixed total number of repeated measurements, we quantitatively compare the efficiency in detecting new intervention effect using DD and SW designs using formulas for compound symmetry covariance structure and empirical calculations for more general Toeplitz correlations. For this we provide insights for researchers in planning longitudinal one-way crossover designs. The following thesis is composed of three chapters represented by three manuscripts. The first chapter develops a unified power estimation approach for continuous outcomes in randomized difference-in-differences (R-DD) studies for both compound symmetry and more general Toeplitz correlation structures that were observed empirically. Optimal number of pre-and post-intervention allocation is analyzed. The second chapter extends the GLS power estimation framework to the non-randomized difference-in-differences (NR-DD) studies and quantitatively compare the penalty of being non-randomized versus randomized for a DD study. Optimal pre-post allocation is also analyzed for NR-DD studies. The third chapter, further investigates on the more general stepped-wedge designs and develop an Orthogonalized Least Squares power estimation framework for both randomized and non-randomized SW (R-SW and NR-SW). The third chapter is research conducted during graduate studies that has been accepted for publication in Statistical Methods in Medical Research published by SAGE.
NotePh.D.
NoteIncludes bibliographical references
Noteby Yirui Hu
Genretheses, ETD doctoral
Persistent URLhttps://doi.org/doi:10.7282/T3Q52RX9
Languageeng
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Version 8.5.5
Rutgers University Libraries - Copyright ©2024