Sample size estimation for correlated count data with changes in dispersion
Jintong Hou
Doctor of Philosophy (Ph.D.), Drexel University
Dec 2024
DOI:
https://doi.org/10.17918/00010855
Files and links (1)
pdf
Hou_Jintong_20241.74 MB
PDF Embargoed Access, Embargo ends: 31 Jan 2027
Abstract
Power analyses for correlated count data are frequently performed in health research. Most of the existing sample size estimation methods for correlated count outcomes were developed under a specified distribution assuming the same distribution or dispersion parameter of the data across measurements and interventions. This assumption may not be flexible in estimation of sample size or power when a change of dispersion happens across measurements in randomized controlled trials or other experimental designs. Existing distributions for correlated count data in sample size estimation include the Poisson distribution, the negative binomial distribution, and the zero-inflated Poisson distribution. In this dissertation, I propose sample size estimation formulas for correlated count data when changes in dispersion occur upon intervention. First, the power and sample size formulas comparing rates before and after an intervention using the rate ratio estimator are derived, assuming the rates before and after an intervention are correlated. Second, power and sample size formulas comparing rate ratios before and after an intervention for two independent interventions are derived, when the change of dispersion occurs after the interventions. Third, empirical power and Type I errors of several estimation methods under small sample size are compared.
Metrics
15 Record Views
Details
Title
Sample size estimation for correlated count data with changes in dispersion
Creators
Jintong Hou
Contributors
Lucy Robinson (Advisor)
Awarding Institution
Drexel University
Degree Awarded
Doctor of Philosophy (Ph.D.)
Publisher
Drexel University; Philadelphia, Pennsylvania
Number of pages
xii, 82 pages
Resource Type
Dissertation
Language
English
Academic Unit
Dana and David Dornsife School of Public Health; Epidemiology and Biostatistics; Drexel University
Other Identifier
991022019220204721
Research Home Page
Browse by research and academic units
Learn about the ETD submission process at Drexel
Learn about the Libraries’ research data management services