Time-Varying Effect Modeling for the Behavioral, Social, and Health Sciences
|Size:||3 MB (3229949 bytes)||Extension:|
|Author(s): Stephanie T. Lanza, Ashley N. Linden-Carmichael
Publisher: Springer, Year: 2021
This book is the first to introduce applied behavioral, social, and health sciences researchers to a new analytic method, the time-varying effect model (TVEM). It details how TVEM may be used to advance research on developmental and dynamic processes by examining how associations between variables change across time. The book describes how TVEM is a direct and intuitive extension of standard linear regression; whereas standard linear regression coefficients are static estimates that do not change with time, TVEM coefficients are allowed to change as continuous functions of real time, including developmental age, historical time, time of day, days since an event, and so forth.
The book introduces readers to new research questions that can be addressed by applying TVEM in their research. Readers gain the practical skills necessary for specifying a wide variety of time-varying effect models, including those with continuous, binary, and count outcomes. The book presents technical details of TVEM estimation and three novel empirical studies focused on developmental questions using TVEM to estimate age-varying effects, historical shifts in behavior and attitudes, and real-time changes across days relative to an event. The volume provides a walkthrough of the process for conducting each of these studies, presenting decisions that were made, and offering sufficient detail so that readers may embark on similar studies in their own research. The book concludes with comments about additional uses of TVEM in applied research as well as software considerations and future directions. Throughout the book, proper interpretation of the output provided by TVEM is emphasized.
Time-Varying Effect Modeling for the Behavioral, Social, and Health Sciences is an essential resource for researchers, clinicians/practitioners as well as graduate students in developmental psychology, public health, statistics and methodology for the social, behavioral, developmental, and public health sciences.
|Table of contents :
Chapter 1: A Conceptual Introduction to Time-Varying Effect Modeling
1.1 What is Time-Varying Effect Modeling?
1.1.1 Time-Invariant Covariates vs. Time-Varying Covariates
1.1.2 Time-Invariant Effects vs. Time-Varying Effects
1.2 TVEM as a Simple Extension of Linear Regression
1.3 Empirical Example: Age-Varying Associations Between Sexual Minority Status and Suicidal Behavior Across Ages 18–60
1.4 Broader Application of TVEM in Social, Behavioral, and Health Research
1.5 Structure of This Book
Chapter 2: Specifying, Estimating, and Interpreting Time-Varying Effect Models
2.1 Data Considerations
2.1.1 Data Coverage Across Time
2.1.2 Types of Data That Can be Analyzed in TVEM
2.1.3 Preparing Data for TVEM
2.2 Estimating Coefficient Functions
2.2.1 Two Approaches to Spline Estimation
184.108.40.206 Model Selection in B-Spline Estimation
2.2.2 Addressing Nonindependence of Repeated Assessments
2.2.3 TVEM Specification in SAS
2.2.4 Weighted Analysis in TVEM
2.3 Model Specification: A Progression Through Four Models
2.3.1 Model 1: Intercept-Only TVEM
2.3.2 Model 2: TVEM With a Main Effect
2.3.3 Model 3: TVEM With a Statistical Control Variable
2.3.4 Model 4: Time-Varying Moderation
2.4 Empirical Example: Age-Varying Association Between Closeness to Mother and Depressive Symptoms
2.4.1 Research Question 1: What is the Mean Level of Depressive Symptoms Across Age in a National Sample of Individuals Followed From Adolescence Through Young Adulthood?
2.4.2 Research Question 2: What is the Age-Varying Association Between Maternal Closeness During Adolescence and Depressive Symptoms Prospectively Through Young Adulthood?
2.4.3 Research Question 3: Does This Age-Varying Association Differ Between Female and Male Individuals?
2.4.4 Sample Results Section
Chapter 3: Generalized Time-Varying Effect Models for Binary and Count Outcomes
3.1 Part I. Generalized TVEM to Model Binary Outcomes
3.1.1 Example: Age-Varying Prevalence of Past-Year Hypertension and Associations With Sex and Racial/Ethnic Group
3.1.2 Research Question 1: What is the Overall Estimated Prevalence of Past-Year Hypertension Across Ages 18–80?
3.1.3 Research Question 2: How Does the Age-Varying Prevalence of Past-Year Hypertension Differ by Sex and by Racial/Ethnic Group? At What Ages are There Significant Group Differences?
220.127.116.11 Model Selection
18.104.22.168 Interpreting Odds Ratio Functions: Calculating Group-Specific Prevalences
3.1.4 Research Question 3: Do Sex and Racial/Ethnic Group Interact to Predict Past-Year Hypertension? (In Other Words, Do Racial/Ethnic Group Differences in Hypertension Across Age Differ by Sex?)
3.1.5 Sample Results Section
3.2 Part II. Generalized TVEM to Model Count Outcomes
3.2.1 Example: Mean Typical Number of Drinks and Associations With Sex and Racial/Ethnic Group Across Age
3.2.2 Research Question 1: What is the Age-Varying Mean Number of Drinks Consumed on a Typical Drinking Occasion in the Past Year, Across Ages 18–35?
3.2.3 Research Question 2: What are the Age-Varying Differences in the Mean Typical Number of Drinks per Drinking Occasion Consumed in the Past Year by Sex and by Racial/Ethnic Group?
22.214.171.124 Differences by Sex
126.96.36.199 Differences Across Racial/Ethnic Groups
3.2.4 Research Question 3: Is there an Interaction Between Sex and Racial/Ethnic Group in Predicting Mean Typical Number of Drinks Consumed per Drinking Occasion in the Past Year?
3.2.5 Sample Results Section
Chapter 4: Time-Varying Effect Modeling to Study Age-Varying Associations
4.1 Examining Differences in Associations Across Age Using TVEM
4.1.1 Research Questions
4.2.3 Statistical Analysis
4.3.1 Research Question 1: What are the Estimated Prevalence Rates of Past-Year Generalized Anxiety Disorder and Past-Year Major Depressive Disorder Across Ages 18–65?
4.3.2 Research Question 2: How Does the Association Between Past-Year MDD and Past-Year GAD Change Across Continuous Age?
4.3.3 Research Question 3: How Does the Association Between Sex and GAD Change Across Continuous Age?
4.3.4 Research Question 4: How Does Sex Moderate, Across Age, the Association Between Past-Year GAD and Past-Year MDD? (In Other Words, How Does the Sex Difference in the Association Between GAD and MDD Differ Across Age?)
Chapter 5: Time-Varying Effect Modeling to Study Historical Change
5.1 Examining Differences in Associations Across Historical Time Using TVEM
5.1.1 Research Questions
5.2.3 Statistical Analysis
5.3.1 Research Question 1: What are the Historical Time Trends From 1990 to 2017 in the Prevalence of High School Seniors Who Perceive Cigarette Smoking as High Risk and in the Prevalence of Seniors Who Report Recent Cigarette Use?
5.3.2 Research Question 2: How Do Associations Between Recent Cigarette Use and (a) Sex and (b) Perceived Risk Associated With Cigarette Smoking Change From 1990 to 2017 Among High School Seniors?
5.3.3 Research Question 3: How Does Sex Moderate, Across Historical Time, the Association Between Perceived Risk of Cigarette Smoking and Recent Cigarette Use?
Chapter 6: Time-Varying Effect Modeling for Intensive Longitudinal Data
6.1 TVEM for Intensive Longitudinal Data
6.2 Current Study
6.3.1 Participants and Procedure
188.8.131.52 Baseline Measures
184.108.40.206 Daily Measures
6.3.3 Statistical Analysis
6.4.1 Research Question 1: How Does the Mean Level of Interparental Conflict Fluctuate Across Days Relative to a Reported Interparental Conflict Event (a) Without Controlling for and (b) While Controlling for Interparental Conflict Events on Other Da
6.4.2 Research Question 2: How Do Associations Between Level of Interparental Conflict and (a) Baseline Family Income, (b) Baseline Interparental Love, and (c) Daily Level of Parent-Child Conflict Vary Across Days Relative to a Reported Interparen
6.4.3 Research Question 3: Do Family Income and Interparental Love Moderate the Association Between Daily Level of Parent-Child Conflict and Daily Level of Interparental Conflict Across Days Relative to a Reported Interparental Conflict Event Afte
Chapter 7: Further Applications and Future Directions
7.1 Comments on Interpreting TVEM Results
7.1.1 Inferring Causation
7.1.2 Lagged Associations and Mediation Analysis
7.1.3 How Do Multilevel Modeling, Growth Curve Modeling, and TVEM Differ?
7.2 Extensions of TVEM and Future Directions
7.2.1 Moving Beyond Time: Other Applications of TVEM
7.2.2 TVEM for Zero-Inflated Count Outcomes
7.2.3 Software, Missing Data, and Sample Size Considerations
7.2.4 Random Effects for Spline Coefficients
7.2.5 MixTVEM: Latent Classes of Individuals Defined by TVEM Coefficients
7.2.6 Time-Varying Coefficients in Latent Class Analysis