In this section, we provide slides for presentations and lectures on a variety of statistical methods in health disparities research. Most are introductory and a few are more advanced.
Testing Mediators and Moderators (Interactions) of Outcomes: Resources
Mediation and moderation are integral methods for investigating health disparities in aging research. We provide several types of resources on these topics including presentations, slides, links to web pages, and PDF documents.
Mediation and Moderation
Mediation and Moderation 101: Tor Neilands (RCMAR Webinar, 2022) A broad overview of historical and current mediation and moderation analysis methods as of 2021. Introduces the concept of statistical mediation and three conceptual approaches to mediation analysis. A YouTube webinar recording, presentation slides, and a Q&A document are available.
Mediation
Introduction to mediation as a concept: David Kenny (2021) Introduces four steps of mediation, indirect effects, power, how to report results, and how to incorporate additional variables. Links to other sites with information on mediation, as well as to his mediation webinars.
Causal mediation analysis in Stata (2024) A webinar introducing basic principles of causal mediation with examples using Stata’s -mediation- command.
Conceptualizing and testing mediated effects: Steve Gregorich (2014) How to conceptualize, test, and interpret a variety of mediation models to explore mechanisms of health disparities.
Moderation
Moderator variables - an introduction: David Kenny (2018) Overview of moderator variables and various situations (e.g., categorical moderators and causal variables) including links to his moderation webinars.
Testing effect modification: Steve Gregorich (2014) Conceptualizing and testing moderated effects: introduces the concept of moderation in the context of exploring mechanisms of health disparities.
Tidbits on exploring interaction effects in logistic regression analyses: Tor Neilands (2024) Methods to decompose and interpret statistically significant interaction effects among categorical and continuous predictors in logistic regression analyses using Stata (with SAS resource links).
Methods of Analyzing Longitudinal Datasets
- Interrupted Time Series Part I Introduction: A gentle introduction to simplified forms of interrupted time series (ITS) analyses (2016).
- Interrupted Time Series Part II Some Analysis Options: Detailed examples of simplified forms of ITS analyses (2016).
- Applications of SAS PROC MIXED: A basic introduction to mixed linear models (2003).
- Applications of Growth Modeling with PROC MIXED: A basic introduction to fitting linear growth models (2001).
- Growth Model Applications: Includes smoothing splines, interrupted times series, and associative latent growth curve models. SAS PROC GLIMMIX offers options to simply specification of smoothing splines (2008).
- Introduction to Mixed Logistic and GEE Models: A basic, older introduction to some frameworks for fitting logistic models to clustered data. More recent advancements have improved some estimation options (2001).
Factor Analysis Methods
- VARCLUS as an Alternative to EFA: An introduction to VARCLUS, a SAS procedure that we (mostly) prefer to exploratory factor analysis (2014).
- Exploratory Factory Analysis: An overview of exploratory factor analysis (2011).
- Testing Measurement Invariance via CFA: A conceptual look at the logic underlying tests of measurement invariance within the confirmatory factor analysis (CFA) framework (2003).
Additional Statistical Issues
- Comparing Odds Ratios Across Nested Logistic Regression Models: Care must be taken when comparing odds ratios across two logistic regression models where the X variables in one model are a subset of the X variables in the other model (2013).
- 3-level Logistic Models: Reviews SAS options circa 2013 for 3-level logistic models, including both multilevel models with random effects and alternating logistic regression (ALR) models. NOTE: SAS now includes the FASTQUAD option in PROC GLIMMIX, which allows fitting models with larger numbers of random effects and quadrature points.
- Missing Data Concepts: A no-math introduction to multiple imputation (2009).