Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


40427.pdf (2.74 MB)

ETD Abstract Container

Abstract Header

Sample Size Determination in Simple Logistic Regression: Formula versus Simulation

Meganathan, Karthikeyan
http://orcid.org/0000-0002-1721-3067
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1627663458916666

Abstract Details

2021, PhD, University of Cincinnati, Medicine: Biostatistics (Environmental Health).
Logistic regression analysis is widely adopted and utilized in several fields. In the medical world, logistic regression lends itself to explain the relationships between risk factors and related clinical events or between treatments and outcomes. Despite its popularity, researchers have not agreed upon a universal approach for sample size determination in logistic regression. Various approaches involving Wald tests, likelihood ratio tests and Score tests have been developed over the last four decades towards this objective. While such approaches simplify calculations of required sample size and related statistical power, the suitability, underlying assumptions, asymptotic approximations and restrictions of these methods are not well recognized. This research compares and contrasts the applicability and performances of 4Wald-based, 2 Likelihood Ratio-based and 2 Score-based methods for sample size determination under a spectrum of significance levels, test powers and model parameters of logistic regression scenarios. Additionally, 2 simplified, simulation-based methods are developed as alternatives for the existing methods. The 8 existing methods all require, in varying forms, knowledge of the logistic model parameters that would identify the underlying joint distribution of the outcome and predictors. Knowledge of the joint distribution facilitates implementing the 2 alternative simulation-based approaches that do not involve assumptions or restrictions. Lastly, the performances of select existing methods and the 2 proposed simulation-based methods are demonstrated using a real example scenario involving nephrology research.
Marepalli Rao, Ph.D. (Committee Chair)
Jane Khoury, Ph.D. (Committee Member)
Anthony Leonard, Ph.D. (Committee Member)
Charuhas Thakar, M.D. (Committee Member)
381 p.
Biostatistics
logistic regression; sample size; simulation; formula

Recommended Citations

Citations

  • Meganathan, K. (2021). Sample Size Determination in Simple Logistic Regression: Formula versus Simulation [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1627663458916666

    APA Style (7th edition)

  • Meganathan, Karthikeyan. Sample Size Determination in Simple Logistic Regression: Formula versus Simulation. 2021. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1627663458916666.

    MLA Style (8th edition)

  • Meganathan, Karthikeyan. "Sample Size Determination in Simple Logistic Regression: Formula versus Simulation." Doctoral dissertation, University of Cincinnati, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1627663458916666

    Chicago Manual of Style (17th edition)

ucin1627663458916666
80
© 2020, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.
Release 3.2.12