Statistical Methods for Global Health and Epidemiology: Principles, Methods and Applications (ICSA Book Series in Statistics)

Preface
Contents
List of Contributors
List of Reviewers
About the Editors
Part I Data Acquisition and Management
	1 Existent Sources of Data for Global Health and Epidemiology
		1.1 Introduction
		1.2 Country Codes, Population and Geographic Area Data
			1.2.1 Standard Country Codes
			1.2.2 Population Data by Country
			1.2.3 Geographic Area Data by Country
			1.2.4 Data from the Internet World Stats
			1.2.5 Data from Wikipedia
		1.3 Data for Socioeconomic Status and Vital Statistics
			1.3.1 Data from the World Health Organization
			1.3.2 Data from the World Bank
		1.4 Data on Important Social, Legal and Religious Factors by Country
			1.4.1 Data for Measuring Press Freedom
			1.4.2 World Index of Moral Freedom
			1.4.3 Country Profile of Religions
		1.5 Data on Disease Statistics
			1.5.1 Data for Global Cancer Statistics
			1.5.2 Data for Global Cardiovascular Disease Statistics
			1.5.3 Data for Global Infectious Disease Statistics
			1.5.4 Data for Causes of Death in the United States
		1.6 Data on Global Tobacco and Substance Use
			1.6.1 Tobacco Use and Prevention
			1.6.2 Alcohol Use
		1.7 Data for Measuring Suicide by Countries in the World
		1.8 Data on Physicians, Nurses and Hospital Beds
		1.9 Important Surveys with International and Global Coverages
			1.9.1 The Demographic and Health Surveys
			1.9.2 Global School-Based Student Health Survey
			1.9.3 Health Behavior in School-Aged Children
			1.9.4 International Social Survey Program
			1.9.5 Multiple Indicator Cluster Survey
			1.9.6 World Health Survey
			1.9.7 The World Mental Health Survey Initiative
			1.9.8 World Value Survey
		1.10 Summary
		References
	2 Satellite Imagery Data for Global Health and Epidemiology
		2.1 Introduction
		2.2 USGS Data
			2.2.1 Introduction of Earth Explorer (EE)
			2.2.2 Steps to Access USGS Data Using EE
		2.3 UNEP Data of United Nations Environmental Program (UNEP)
			2.3.1 Introduction of the Environmental Data Explorer of UNEP
			2.3.2 Steps to Access UNEP Data
		2.4 NASA Earth Science Data
			2.4.1 Introduction of the Earth Science Data
			2.4.2 Steps to Access Earth Science Data
		2.5 Sentinel Satellite Data
			2.5.1 Introduction of the Sentinel Satellite Data
			2.5.2 Steps to Access the Sentinel Satellite Data
		2.6 Global ALOS 3D World Data
			2.6.1 Introduction of the Global ALOS 3D World
			2.6.2 Steps to Access the ALOS 3D World Data
		2.7 Earth Online Data
			2.7.1 Introduction to the EO Data
			2.7.2 The Steps of Access Earth Online Data
		2.8 Additional Sources of Data
			2.8.1 Comprehensive Large Array-Data Stewardship System (CLASS)
			2.8.2 National Institute for Space Research (INPE)
			2.8.3 Himawari Monitor Data of Japanese Meteorological Agency (JMA)
			2.8.4 The AErosol RObotic NETwork (AERONET)
			2.8.5 Bhuvan India Geo-Platform of ISRO
		2.9 Conclusion Remark
		References
	3 GIS/GPS-Assisted Probability Sampling in Resource-Limited Settings
		3.1 Study Population and Samples
		3.2 Non-probability Sampling
			3.2.1 Purposeful Sampling
			3.2.2 Convenience Sampling
		3.3 Probability Sampling
			3.3.1 Know the Probability for Sampling
			3.3.2 Independent Identical Sample Distribution
			3.3.3 Generalizability to the Study Population
		3.4 Challenges to the Classic Probability Sampling Methods and Alternatives
			3.4.1 Methodology Barriers
			3.4.2 Hard-to-Reach or Hidden Populations
			3.4.3 Urgency to Know Study Results
			3.4.4 Application of GIS/GPS Technologies in Probability Sampling
		3.5 Challenges to the Existing GIS/GPS-Assisted Probability Sampling Methods
			3.5.1 Challenges to Determine Sample Size Before Sampling
			3.5.2 Challenges to Distinguishing Residential from Non-residential Housing
			3.5.3 Challenges Due to Heterogeneity in Population Density
			3.5.4 Challenges to Determine the Geographic Sample Weights
		3.6 GIS/GPS Assisted Multi-stage Probability Sampling
			3.6.1 Introduction to the Method
			3.6.2 Stage 1 Sampling: Random Selection of Geographic Units
			3.6.3 Stage 2 and 3: Random Selection of Geographic Segments and Households
			3.6.4 Stage 4: Random Selection of Participants from Households
			3.6.5 Complementary Data Collection
		3.7 Methods to Determine Residential Areas
			3.7.1 Method 1. Estimate Residential Area with Collected Data
			3.7.2 Method 2. Estimate Residential Area with Monte Carlo Method
		3.8 Estimate of Sample Weights
		3.9 Practical Test of the Method in an NIH Funded Project
			3.9.1 Geographic Sampling Frame and Geounits
			3.9.2 Sampling Geographic Segments, Households and Participants
			3.9.3 Determination of Residential Area with Imagery Data and GPS-Tracking File
		3.10 Strengths and Recommendation
			3.10.1 Strengths
			3.10.2 Recommendations for Application
		Appendix 1: R Program Codes for a Semi-automatic, Computer-Assisted, and Step-Wise Algorithm for Geounit Sampling
		Appendix 2: R Program Codes for Monte Carlo Method to Determine Residential Area
		References
	4 Construal Level Theory Supported Method for Sensitive Topics: Applications in Three Different Populations
		4.1 Introduction
			4.1.1 Factors Affecting the Quality of Survey Data
			4.1.2 Cognitive Censoring and Social Desirability Bias
			4.1.3 Existing Methods to Reduce Social Desirability Bias
		4.2 Theoretical and Analytical Foundations
			4.2.1 Construal-Level Theory and Social Desirability Bias
			4.2.2 The Measurement Theory Underpinning of CLT-Based Survey
			4.2.3 Statistical Modeling of CLT-Based Survey Data
			4.2.4 Bifactor and Tri-factor Modeling Analysis of CLT-Based Data
		4.3 Detecting the Sensitive of a Question Using CLT-Based Method
			4.3.1 Participants and Procedures
			4.3.2 Statistical Analysis and Results
			4.3.3 Summary
		4.4 Application of CLT-Based Method in an Urban Population
			4.4.1 Participants and Procedures
			4.4.2 Conventional and CLT-Based Brief Sexual Openness Scale
			4.4.3 Variable for Predictive Validity Analysis
			4.4.4 Statistical Analysis
			4.4.5 Sample Characteristics
			4.4.6 Performance of the BSOS as a Conventional Scale
			4.4.7 Performance of the CLT-Based Method for Assessing Single Questions
			4.4.8 Construct Validity of CLT-Based Method as a Multi-contents Instrument
			4.4.9 Separation of Three Factors Based on CLT-Based Data
			4.4.10 Bias Assessment
			4.4.11 Predictive Validity
			4.4.12 Summary
		4.5 Application of the CLT-Based Method in an Rural Sample
			4.5.1 Data Sources and Participants
			4.5.2 BSOS as Conventional and CLT-Based Scale
			4.5.3 Variables for Validity Assessment
			4.5.4 Statistical Analysis
			4.5.5 Sample Characteristics
			4.5.6 Results from Tri-factor Analysis
			4.5.7 Predictive Validity
			4.5.8 Summary
		4.6 Discussion and Conclusions
			4.6.1 Theoretical Framework of the CLT-Based Method
			4.6.2 Empirical Support for the CLT-Based Method
			4.6.3 Recommendations and Future Research
		References
	5 Integrative Data Analysis and the Study of Global Health
		5.1 Pooled Data Analysis and Global Health Research
		5.2 Integrative Data Analysis
			5.2.1 Defining Integrative Data Analysis
			5.2.2 Research Questions Suitable for IDA
		5.3 Measurement Harmonization
			5.3.1 Need for Measurement Harmonization
			5.3.2 Logical Harmonization
		5.4 Harmonization
			5.4.1 Psychometric Harmonization and IDA
			5.4.2 Psychometric Harmonization Model
		5.5 Illustrative Example
			5.5.1 Logical Harmonization of Individual Items
			5.5.2 Steps 1 and 2: Descriptive Analysis
			5.5.3 Step 3: Iterative MNLFA
			5.5.4 Step 4: Examine MNLFA Scores
		5.6 Hypothesis Testing in IDA
			5.6.1 Challenges to Hypothesis Testing
			5.6.2 Another Example of IDA in Global Health Research
		5.7 Advances in IDA Methods
			5.7.1 New Regularization Method to Identify DIF Items
			5.7.2 Trifactor Modeling Method
			5.7.3 Summary and Conclusions
		Technical Appendix: Moderated Nonlinear Factor Analysis (MNLFA)
			Initial Models
			Simultaneous Model
			Final Model
		References
	6 Introduction to Privacy-Preserving Data Collection and Sharing Methods for Global Health Research
		6.1 Introduction
		6.2 Randomized Response Technique and Its Extensions
		6.3 Warner\'s Method
			6.3.1 Principles and Method
			6.3.2 Application of Warner\'s Method in Study Risky Behaviors Among College Students
			6.3.3 Limitations of Warner\'s Method
				6.3.3.1 Other Extensions of Warner\'s Method
		6.4 More Sophisticated Randomized Response Techniques: RAPPOR
		6.5 Random Orthogonal Matrix Masking (ROMM)for Data Sharing
			6.5.1 Basic Principles and Methodology
			6.5.2 Examples of Random Transformation
		6.6 Triple Matrix-Masking (TM2) Methods
			6.6.1 Principles and Methodology
			6.6.2 Extensions of TM2 Methods
		6.7 Conclusion Remarks
		References
Part II Essential Statistical Methods
	7 Geographic Mapping for Global Health Research
		7.1 Importance of Global Mapping
		7.2 Preparation for Geographic Mapping
			7.2.1 Brief Introduction to R and R Studio
			7.2.2 Download and Install R
			7.2.3 Download and Install R Studio
			7.2.4 Work Around R Studio
		7.3 R Packages for Geographic Mapping
			7.3.1 R Packages Needed
			7.3.2 Download and Install the Related R Packages
		7.4 Mapping the World Using R
			7.4.1 Creating a Base World Map
			7.4.2 Change Map Projections for Best View
			7.4.3 Map Rotation for a Different Central View
			7.4.4 An Example with Both Projection and Rotation
		7.5 Geographic Mapping of the World Population: A Practical Example
			7.5.1 Steps to Map a Subject Matter
			7.5.2 Data Preparation
			7.5.3 Mapping Your Data
		7.6 Mapping the Density of World Population by Country
		7.7 Conclusion Remarks
		Appendix
		References
	8 A 4D Indicator System of Count, P Rate, G Rate and PG Rate for Epidemiology and Global Health
		8.1 Introduction
		8.2 Ending the HIV/AIDS Epidemic by 2030
		8.3 Four-Dimensional Measurement System
			8.3.1 Two Conventional Measure of Headcount and P Rate
			8.3.2 Two New Measures of G Rate and P Rate
		8.4 An Example of Global HIV Epidemic
			8.4.1 Materials and Method
			8.4.2 Estimation of P Rate, G Rate and PG Rate
			8.4.3 Geographic Mapping
		8.5 Results
			8.5.1 The Global HIV Epidemic Measured by Headcounts of PLWH
			8.5.2 The Global HIV Epidemic Measured by P Rates of PLWH
			8.5.3 The Global HIV Epidemic Measured by G Rates of PLWH
			8.5.4 The Global HIV Epidemic Measured by PG Rates of PLWH
		8.6 Discussion and Conclusion Remarks
		A.1 Appendix 1. List of countries with population, land area, total PLWH, P rate, G rate and PG rate
		References
	9 Historical Trends in Mortality Risk over 100-Year Period in China with Recent Data: An Innovative Application of Age-Period-Cohort Modeling
		9.1 Introduction
			9.1.1 Learn from History
			9.1.2 Challenges for Quantitative Historical Research
		9.2 Timeline of Significant Events in China Since 1900
			9.2.1 Overthrow of the Feudalistic Society
			9.2.2 Early Period After Independence
			9.2.3 The Period of Open Policy and Economic Reform
		9.3 Age-Specific Data and APC Modeling Analysis
			9.3.1 Age-Specific Data as Digital Fossils
			9.3.2 APC Model to Extract the Historical Information
			9.3.3 Challenges to APC Modeling
			9.3.4 New Data Selection Method to Correctly Estimate Cohort Effect
			9.3.5 Using Single Year of Data 5 Years Apart as a Solution
		9.4 Materials and Methods
			9.4.1 Source of Data
			9.4.2 APC Modeling Analysis
		9.5 Main Study Findings
			9.5.1 Visual Presentation of the Mortality Data
			9.5.2 Comparison of Results from Four Different APC Models
			9.5.3 Period Effect for Mortality Risk Change over 1990–2010
			9.5.4 Changes in Cohort Effect Through Numerical Differentiation
			9.5.5 Sunny Periods in Historical China
			9.5.6 Cloudy Period in Historical China
		9.6 Discussion and Conclusions
			9.6.1 Findings and Implications for China
			9.6.2 Economic and Technic Advancement Not Equal to Good Health
			9.6.3 APC Modeling for Historical Epidemiology
			9.6.4 Implication for Research in Other Countries
			9.6.5 Limitations and Conclusion Remarks
		References
	10 Moore-Penrose Generalized-Inverse Solution to APC Modeling for Historical Epidemiology and Global Health
		10.1 Introduction
		10.2 APC Model and Its Estimation
			10.2.1 An Introduction to APC Model
			10.2.2 Solving APC Using MP Method
		10.3 Application with Real Data
			10.3.1 Data Source and Arrangement
			10.3.2 Modeling Analysis with MP-APC
			10.3.3 Comparison with Results from IE-APC
		10.4 Relationship Between IE-APC and MP-APC
			10.4.1 IE-APC Modeling
			10.4.2 MP-APC Modeling
		10.5 Discussion and Conclusions
		Appendix: R Program for MP-APC
		References
	11 Mixed Effects Modeling of Multi-site Data-Health Behaviors Among Adolescents in Hong Kong, Macao, Taipei, Wuhan and Zhuhai
		11.1 Introduction
		11.2 Methodology Challenge and Alternatives
			11.2.1 Heterogeneity Data for Global Health Research
			11.2.2 Understand Multi-site and Multi-level Data
		11.3 A Study Across Five Chinese Cities: An Example
			11.3.1 Purposes and Rational
			11.3.2 Participants and Procedure
			11.3.3 Measurement of Lifestyle Behavior
			11.3.4 Measurement of Addictive Behaviors
			11.3.5 Measurement of Student-Level Factors
			11.3.6 Measurement of Site-Level Factors
		11.4 Statistical Analysis and Results
			11.4.1 Data Analysis
			11.4.2 Study Site and Sample
			11.4.3 Prevalence of Life Style Variables
			11.4.4 Prevalence of Addictive Behaviors
			11.4.5 Intraclass Correlation for the Variable Time on Siting Position
			11.4.6 Results from Mixed Effects Model and Linear Regression
		11.5 Discussion and Conclusions
			11.5.1 Significance of the Mixed Effects Modeling Methods
			11.5.2 Implications of the Findings from This Study
		References
	12 Geographically Weighted Regression
		12.1 Introduction
		12.2 Theory
			12.2.1 Basic Model Structure and Inference
			12.2.2 Constructing Weights
			12.2.3 Testing Spatial Nonstationarity
			12.2.4 Geographically Weighted Generalized Linear Models
			12.2.5 Colinearity and Remedies
				Local Linear Estimation
				Regularized Fitting
		12.3 Software and Case Study
			12.3.1 Data
			12.3.2 Data Analysis with R Packages
			12.3.3 Conclusion
		References
Part III Advanced Statistical Methods
	13 Bayesian Spatial-Temporal Disease Modeling with Application to Malaria
		13.1 Introduction
		13.2 Spatial-Temporal Data in Nigeria
			13.2.1 Study Area
			13.2.2 Country Profile
			13.2.3 Ethical Approval
			13.2.4 Predictor Variables
		13.3 Statistical Methodology
			13.3.1 Malaria Spatial-Temporal Modeling
				Data Distribution
				Spatial-Temporal Mixed-Effects Regression Model
		13.4 Bayesian Spatial-Temporal Models with INLA
			13.4.1 Goodness of Fit Statistics
		13.5 Results
		13.6 Conclusion and Summary of Findings
		A.1 Appendix 1: R Program Codes for Analysis.
		References
	14 BCEWMA: A New and Effective Biosurveillance System for Disease Outbreak Detection
		14.1 Introduction
		14.2 Some Basic SPC Concepts and Methods
		14.3 A New Biosurveillance System
			14.3.1 A Baseline Model and Its Estimation
			14.3.2 Sequential Monitoring of Disease Incidence Rates
		14.4 Real Data Examples
			14.4.1 The Hand, Foot and Mouth Disease Data
			14.4.2 The Influenza-Like-Illness Data
		14.5 Concluding Remarks
		References
	15 Cusp Catastrophe Regression Analysis of Testosterone in Bifurcating the Age-Related Changes in PSA, a Biomarker for Prostate Cancer
		15.1 Introduction
			15.1.1 Challenges to Using PSA as Prostate Cancer Screener
			15.1.2 Age Pattern of PSA Changes
			15.1.3 Relationship Between Testosterone and PSA
			15.1.4 A Cusp Catastrophe Model of PSA as Function of Age and Testosterone
			15.1.5 Purpose of This Study
		15.2 Materials
			15.2.1 Participants and Data
			15.2.2 Variables and Measurement
		15.3 Statistical Analysis and Cusp Modeling
			15.3.1 Statistical Analysis
			15.3.2 Cusp Catastrophe Modeling
		15.4 Analytical Findings
			15.4.1 Sample Characteristics
			15.4.2 Results from Linear Correlation Analysis
			15.4.3 Results from Linear Regression Modeling
			15.4.4 Bimodality of the PSA Level in Men
			15.4.5 Results from Cobb-Grasman Cusp Modeling
			15.4.6 Results from Chen-Chen Cusp Regression Modeling
			15.4.7 Cusp Point, Threshold Lines and Cusp Region
		15.5 Discussion and Conclusions
			15.5.1 PSA Dynamics Is Nonlinear and Discrete
			15.5.2 Co-use of Testosterone and PSA for Screening
			15.5.3 Limitations and Future Research
		References
	16 Logistic Cusp Catastrophe Regression for Binary Outcome: Method Development and Empirical Testing
		16.1 Background
			16.1.1 Cusp Catastrophe for Nonlinear Discrete Systems
			16.1.2 Established Methods for Cusp Catastrophe Modeling
			16.1.3 Need for Methods to Model Binary Data
		16.2 An Overview of the Cusp Catastrophe Model
			16.2.1 Deterministic Cusp Model
			16.2.2 Characteristics of the Cusp Catastrophe Model
		16.3 Implementation of a Cusp Catastrophe Model
			16.3.1 Guastello\'s Polynomial Approach
			16.3.2 Cobb-Grasman\'s Approach
			16.3.3 Chen-Chen\'s Cusp Regression Approach
		16.4 Cusp Catastrophe Modeling of Binary Data
			16.4.1 The Binary Data Structure
			16.4.2 The Binary Cusp Catastrophe Model
			16.4.3 Maximun Likelihood Estimation
			16.4.4 Cusp Catastrophe Conventions
			16.4.5 Cusp Region Estimation
			16.4.6 Numeric Search Algorithms for Parameter Estimates
		16.5 Test the Logistic Cusp Catastrophe Model Through Monte-Carlo Simualtion
			16.5.1 Model Settings for Simulation
			16.5.2 Steps of Simulation Study
			16.5.3 Results and Interpretation
		16.6 Modeling Analysis with Real Data: Binge Drinking
			16.6.1 Data Sources and Variables
			16.6.2 Modeling Analysis
			16.6.3 Parameter Estimates and Comparison
			16.6.4 Comparison of the Estimated Cusp Regions
		16.7 Discussion and Conclusions
		References
Correction to: Statistical Methods for Global Health and Epidemiology
Index