Material Type  

 학위논문

 

0016933636

Date and Time of Latest Transaction  

20240214101316

ISBN  

9798379565008

DDC  

310

Author  

Wakim, Nicky Irene.

Title/Author  

Methods for Summarizing and Imputing High-Frequency Digital Biomarkers in the Context of Longitudinal Data Analysis - [electronic resource]

Publish Info  

[S.l.] : University of Michigan., 2023

Publish Info  

Ann Arbor : ProQuest Dissertations & Theses, 2023

Material Info  

1 online resource(121 p.)

General Note  

Source: Dissertations Abstracts International, Volume: 84-12, Section: B.

General Note  

Advisor: Braun, Tom;Wu, Zhenke.

학위논문주기  

Thesis (Ph.D.)--University of Michigan, 2023.

Restrictions on Access Note  

This item must not be sold to any third party vendors.

Restrictions on Access Note  

This item must not be added to any third party search indexes.

Abstracts/Etc  

요약Digital biomarkers have the potential to aid early detection of cognitive decline, resulting in interventions that delay dementia onset. Digital biomarkers are measured at high frequencies, which increases opportunities for assessment of cognitive ability. Summarization of digital biomarkers helps to reduce computational burden or variance among values. However, the consequences of summarization processes are not statistically investigated, especially the potential bias in longitudinal data analysis. This dissertation will systematically investigate methods for summarization and imputation of high-frequency digital biomarker data. These methods are created in the context of longitudinal analysis of a binary cognitive outcome with digital biomarker predictors. In Chapter 2, we discuss factors of high-frequency digital biomarker data that need to be considered when faced with a time granularity decision, defined as the frequency at which measurements are observed or summarized. It is important to find a balance between ease of analysis by condensing data and the integrity of the data, which is reflected in a chosen time granularity. Via simulation, we investigate the factors and examine how each affects the ability to detect the true, underlying digital biomarker pattern using root mean squared error. Then we apply our procedure to the Intelligent Systems for Assessing Aging Change (ISAAC) study, which involves longitudinal walking speed. The example sheds light on typical problems data present and how we can use the above factors in exploratory analysis to choose an appropriate time granularity. In Chapter 3, we examine the process of simultaneously imputing and summarizing digital biomarker data. When analyzing the association of digital biomarker data and a lower-frequency outcome, we want to summarize biomarker data to match the outcome's frequency while simultaneously imputing missing biomarker values. We define two methods for the imputation and summarization process: (1) Impute then Summarize, where we impute at the finest time granularity, then summarize, and (2) Summarize then Impute, where we summarize the biomarker then impute at the summary level. Via simulation, we assess two processes involving imputation of biomarkers for longitudinal analysis of a binary outcome. Our results show that accuracy of coefficient estimation depends on percent missing data, length of consecutive missing days, and the rate of trajectory change of the biomarkers. We apply these processes to the ISAAC study, and find that the odds of testing for mild cognitive impairment is negatively associated with increased walking speed. In Chapter 4, we investigate simultaneous imputation of multiple, potentially correlated digital biomarkers, and association analysis of one biomarker with a longitudinal outcome. We build off the two methods in Chapter 3 using an updated regression model that incorporates information from additional biomarkers through random effects. Via simulation, we vary levels of correlation between digital biomarkers and percent missing values of each digital biomarker to examine the effect on imputed digital biomarker values and relative bias of coefficient estimates from longitudinal analysis. the coefficient estimate of the longitudinal analysis. Our results show increased correlation has the biggest effect on reducing relative bias, followed by increased number of digital biomarkers. We also found that at low correlation levels, an increase in percent missing values of other digital biomarkers results in the decreased in magnitude of relative bias. We apply our imputation methods to the ISAAC study with two and four digital biomarkers, and find that imputed values do not improve with additional digital biomarkers.

Subject Added Entry-Topical Term  

Statistics.

Subject Added Entry-Topical Term  

Biostatistics.

Index Term-Uncontrolled  

Missing data

Index Term-Uncontrolled  

Digital biomarkers

Index Term-Uncontrolled  

Imputations

Index Term-Uncontrolled  

High-frequency predictors

Index Term-Uncontrolled  

Longitudinal data analysis

Added Entry-Corporate Name  

University of Michigan Biostatistics

Host Item Entry  

Dissertations Abstracts International. 84-12B.

Host Item Entry  

Dissertation Abstract International

Electronic Location and Access  

로그인을 한후 보실 수 있는 자료입니다.

소장사항  

202402 2024

Control Number  

joongbu:641003