Statistical Analysis of Human Growth and Development
Chapman and Hall/CRC – 2013 – 378 pages
Statistical Analysis of Human Growth and Development is an accessible and practical guide to a wide range of basic and advanced statistical methods that are useful for studying human growth and development. Designed for nonstatisticians and statisticians new to the analysis of growth and development data, the book collects methods scattered throughout the literature and explains how to use them to solve common research problems. It also discusses how well a method addresses a specific scientific question and how to interpret and present the analytic results. Stata is used to implement the analyses, with Stata codes and macros for generating example data sets, a detrended Q-Q plot, and weighted maximum likelihood estimation of binary items available on the book’s CRC Press web page.
After reviewing research designs and basic statistical tools, the author discusses the use of existing tools to transform raw data into analyzable variables and back-transform them to raw data. He covers regression analysis of quantitative, binary, and censored data as well as the analysis of repeated measurements and clustered data. He also describes the development of new growth references and developmental indices, the generation of key variables based on longitudinal data, and the processes to verify the validity and reliability of measurement tools. Looking at the larger picture of research practice, the book concludes with coverage of missing values, multiplicity problems, and multivariable regression.
Along with two simulated data sets, numerous examples from real experimental and observational studies illustrate the concepts and methods. Although the book focuses on examples of anthropometric measurements and changes in cognitive, social-emotional, locomotor, and other abilities, the ideas are applicable to many other physical and psychosocial phenomena, such as lung function and depressive symptoms.
"… a very useful and well-organised manual to be used by statisticians and non-statisticians. It responds to the problems that may arise on both sides."
—Journal of Applied Statistics, 2014
"… a very useful overview of the vast array of statistical tools, models and tests, which can be applied to the study of human growth and development and also to some important concepts, which are relevant to epidemiological research more generally. The code provided makes this book particularly accessible to those who are familiar with Stata or who are intending to use this software for their analysis, although the methods described could easily be implemented in other standard statistical software packages."
—Corrie Macdonald-Wallis, International Journal of Epidemiology, 2014
Causal Reasoning and Study Designs
Basic Statistical Concepts and Tools
Statistical Inference and Significance
Quantifying Growth and Development: Use of Existing Tools
Regression Analysis of Quantitative Outcomes
Regression Analysis of Binary Outcomes
Introduction to Generalized Linear Models
Log-Binomial and Binomial Regression Models
Regression Analysis of Censored Outcomes
Regression Analysis of Right-Censored Data
Analysis of Interval-Censored Data
Analysis of Repeated Measurements and Clustered Data
Robust Variance Estimator
Analysis of Subject-Level Summary Statistics
Quantifying Growth: Development of New Tools
Capturing Nonlinear Relationships
Quantifying Development: Development of New Tools
Summary Index Based on Binary Items
Summary Index Based on Quantitative Variables
Defining Growth and Development: Longitudinal Measurements
Expected Change and Unexplained Residuals
Reference Intervals for Longitudinal Monitoring
Conditional Scores by Quantile Regression
Validity and Reliability
Missing Values and Imputation
When Is Missing Data (Not) a Problem?
Interpolation and Extrapolation
Imputing Censored Data
When Not to Do Multiplicity Adjustment
Strategies of Analysis to Prevent Multiplicity
Close Testing Procedure
Regression Analysis Strategy
Rationale of Using Multivariable Regression
Point Measures, Change Scores, and Unexplained Residuals
Issues in Variable Selection
Role of Prior Knowledge
Appendix A: Stata Codes to Generate Simulated Clinical Trial (SCT) Dataset
Appendix B: Stata Codes to Generate Simulated Longitudinal Study (SLS) Dataset
Appendix C: Stata Program for Detrended Q-Q Plot
Appendix D: Weighted Maximum Likelihood Estimation for Binary Items
Yin Bun Cheung is a professor in the Centre for Quantitative Medicine, Office of Clinical Sciences at Duke-NUS Graduate Medical School in Singapore and adjunct professor in the Department of International Health at the University of Tampere in Finland. Dr. Cheung has been studying human growth and development in African and Asian countries for about 15 years.