Development of Agreement Measures for Studying Consistency of Menstrual Cycle Length and its Relationship to Fertility Open Access

Daya, Natalie Rula (2012)

Permanent URL: https://etd.library.emory.edu/concern/etds/8336h2093?locale=en
Published

Abstract

The menstrual cycle acts as an overt indicator of reproductive health. Irregular
menstruation with inconsistency in cycle length may affect fertility and risk of chronic
diseases such as breast cancer. In this study we are interested in quantifying the
consistency of menstrual cycle length in terms of an agreement measure and evaluating
the association between the consistency of cycle length and a woman's reproductive
health and characteristics such as diet and lifestyle. There are several challenges in
conducting agreement analysis for our study: there is an unequal number of cycle length
observations across subjects; there are censored observations in recorded cycle lengths;
standard agreement method only provides group-level summary statistics and does not
allow modeling agreement in terms of subject-specific covariates. In this thesis, we aim
to develop appropriate statistical methods to address these issues. To accommodate
censoring, we propose to impute cycle lengths for censored observations using the mean
residual life estimated based on a parametric model. We first consider a group-level
agreement method based on the intra-class correlation coefficient. We then propose a
novel within-subject agreement method for replicated measurements by generating paired
data based on the replications within each subject and then estimating an agreement
measure based on the generated pairs. We consider subject-specific agreement index
based on two different measures: Concordance Correlation Coefficient (CCC) and Total
Deviation Index (TDI) and compare their performance. We then model the subject-
specific agreement measures in terms of subject's covariates effects. Our results show
that the new within-subject agreement method provides more biologically meaningful
results than the group-level method. The TDI which is a more straightforward agreement
measure provides better results than the CCC which is a scaled agreement measure.
Consistency of a woman's menstrual cycle length is significantly associated with
pregnancy status, stress level and caffeine consumption.

Table of Contents

List of Tables ........................................................................................................................ i
List of Figures ...................................................................................................................... ii


1. Introduction.....................................................................................................................1


2. Motivation and Background...............................................................................................5
2.1 Motivation example ...........................................................................................................5
2.2 Some agreement measures .................................................................................................6


3. Descriptive statistics of the data set..................................................................................9


4. Accommodation of censored observations........................................................................10


5. Group-level agreement method........................................................................................12


6. Subject-specific agreement method.................................................................................15
6.1 Generation of paired-observations based on replicated measurements .......................................15
6.2 Concordance Correlation Coefficient ...................................................................................15
6.3 Total Deviation Index .......................................................................................................18


7. Modeling.........................................................................................................................24
7.1 Covariates .....................................................................................................................24
7.2 Modeling covariates .........................................................................................................27
7.3 Sensitivity analysis...........................................................................................................31


8. Conclusions....................................................................................................................32
8.1 Summary .......................................................................................................................32
8.2 Future research ..............................................................................................................34


References ..........................................................................................................................35
Appendix .............................................................................................................................36

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files