Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List

Full text release has been delayed at the author's request until December 10, 2024

ETD Abstract Container

Abstract Header

A Novel Count Weighted Wilcoxon Rank-Sum Test and Application to Medical Data

Cong, Xinyu
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1668637573266784

Abstract Details

2022, PhD, University of Cincinnati, Medicine: Biostatistics (Environmental Health).
Cortical spreading depolarizations (SDs) are abnormal activities in the brain observed after traumatic injury. These electrophysiologic dysfunctions occur in many patients with severe traumatic brain injury (TBI) during acute neurocritical care in the form of short-circuiting waves that can last from periods of minutes to hours. In chapter 2 of this dissertation, we use data obtained from electrocorticographic (ECoG) monitoring of depolarization activities of patients who underwent neurosurgical treatment for TBI. We first review and apply statistical methods and approaches commonly used in the analysis of electrophysiologic studies of traumatic brain injury. This includes Wilcoxon rank-sum tests for continuous variables to compare medians, and linear and generalized linear regression models to investigate connec- tions between multiple parameters and SD events. We also build and explore multivariate models for predicting SD events. During neurointensive care, different physiological variables of patients could be recorded at various regular temporal intervals. While some variables are recorded at very small-time intervals (e.g., every minute), some may have large recording intervals (e.g., 3 hours). It is also possible that monitoring windows for different variables may have different frequencies with gaps without any measurements. Thus, when we focus on investigating links between more than two variables, aligning hourly data will drastically decrease our sample size. Therefore, to solve the problem of temporal misalignment while trying to use as much information in the original data as possible, the third chapter of the dissertation explores various temporal interpolation methods to impute missing data created from data alignment. We first explore the possibility of applying spatial statistics to the imputations and then compare it with two simple imputation methods: linear interpolation and spline-based interpolation. It turns out that linear interpolation is easy to understand and apply but only imputes a small amount of missing data. The spline-based interpolation can impute all missing data, but the results may be inaccurate. Surprisingly, the spatial method can impute almost all missing data and result in a solid accuracy. In chapter 4, we propose a modification to the Wilcoxon rank-sum (WMW) test as an attempt to increase the detection power of the approach by making it applicable to count data. The basic WMW test is a non-parametric test that allows for comparisons of characteristics between two patient populations, and therefore can easily be utilized to analyze associations between binary outcomes and the parameters of interest. However, when the binary variable becomes a count variable, the conventional WMW test will not be directly applicable. One way to continue the WMW test is to dichotomize the count variable. However, doing so will lose information. We propose a count weighted WMW test in chapter 4. The new method allows us to take into consideration the effect of count and hence can potentially increase the detection power of the test. We then compare the count weighted method to the conventional non-parametric tests using simulated and real data and conclude that the count weighted method surpasses the conventional non-parametric test.
Roman Jandarov, Ph.D. (Committee Member)
Marepalli Rao, Ph.D. (Committee Member)
Jed Hartings, Ph.D. (Committee Member)
59 p.
Biostatistics
non-parametric; spatial statistics; spreading depolarization; count

Recommended Citations

Citations

  • Cong, X. (2022). A Novel Count Weighted Wilcoxon Rank-Sum Test and Application to Medical Data [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1668637573266784

    APA Style (7th edition)

  • Cong, Xinyu. A Novel Count Weighted Wilcoxon Rank-Sum Test and Application to Medical Data. 2022. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1668637573266784.

    MLA Style (8th edition)

  • Cong, Xinyu. "A Novel Count Weighted Wilcoxon Rank-Sum Test and Application to Medical Data." Doctoral dissertation, University of Cincinnati, 2022. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1668637573266784

    Chicago Manual of Style (17th edition)

ucin1668637573266784
© 2022, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.