Mediation Analysis

Mediation analysis is used widely throughout disciplines that rely on behavioral science methodologies. In this class we address statistical methods used to test hypotheses about the mechanisms by which causal effects operate. Topics covered include:

• Partitioning a variable X's total effect on Y into its components direct and indirect through a mediator M using ordinary least squares regression.
• Inference about pathways of influence of X's effect, including the use of modern bootstrap-based methods for inference about indirect effects.
• Estimation and interpretation of mediation models with more than one mediator.
• The application of mediation analysis to designs in which the causal antecedent X is multicategorical (three or more categories).
• Mediation analysis in the two group pretest-posttest design.
• Debiasing the estimation of effects by accounting for random measurement error in causes and mediators.
• Use of the PROCESS macro for SPSS, SAS, and R to simplify an analysis relative to using alternative analytical approaches.

This course meets in person in Calgary July 15-16, 2025, between 9am and 5pm each day.

Computer applications will focus on the use of ordinary least squares regression and the PROCESS macro for SPSS, SAS and R, developed by the instructor, that makes the analyses described in this class much easier than they otherwise would be. This is a hands-on course, so maximum benefit results when learners can follow along with analyses using a laptop or desktop computer with a recent version of SPSS Statistics (version 27 or later), SAS (release 9.3 or later, with PROC IML installed) or R (version 3.6 or later; base module only. No packages are used in this course). Learners can choose which statistical package they prefer to use. STATA users can benefit from the course content, but PROCESS makes these analyses much easier and is not available for STATA. 

This course will be helpful for researchers in any field – including psychology, sociology, education, business, human development, social work, public health, communication and others that rely on social science methodology – who want to learn how to apply the methods of mediation analysis using widely-used software such as SPSS, SAS and R.

Learners are recommended to have familiarity with the practice of multiple regression analysis and elementary statistical inference. No knowledge of matrix algebra is required or assumed, nor is matrix algebra used in the delivery of course content. Learners should also have some experience with the use of SPSS, SAS or R, including opening and executing data files and programs.

By the end of this course, you will…

• be able to statistically partition one variable’s total effect on another into its primary pathways of influence, direct and indirect.
• understand historical and modern approaches to inference about indirect effects in causal models, including bootstrapping and related modern methods.
• know how test competing theories of mechanisms statistically through the comparison of indirect effects in models with multiple mediators
• understand the application and interpretation of mediation analysis in models that include multicategorical causes (such as three conditions in an experiment) and the distinction between relative total, direct, and indirect effects.
• be able to describe and understand the differences between traditional and counterfactual approaches to mediation analysis.
• have developed the skill to discuss and choose between mediation analysis models in the two-instance experimental and longitudinal pretest-posttest design. 
• have acquired the ability to discuss and understand the advantages and disadvantages of regression-based and structural equation modeling-based approaches to mediation analysis
• know how to conduct a mediation analysis using readily-available statistical software, including the PROCESS procedure for SPSS, SAS, and R.
• be prepared to advance, on your own time in the future, your knowledge of mediation analysis developed in this class to more complex designs and when using measurement procedures that don’t yield continuous data.
  
   A certificate of completion from the Canadian Centre for Research Analysis and Methods is provided at the end of the course.

Statistical mediation analysis is a vast topic, with the specifics of various methods dependent on design and measurement. This course is designed to give you a solid introduction to the fundamentals with a focus on modeling continuous outcomes using regression analysis approaches and tools. Some of the more complex topics that are not covered in this course include multilevel models and designs; mediation analysis with dichotomous, count, or ordinal outcomes; survival models. But once you complete this course, you will be better prepared to dive deeper on your own into the literature on these more complex design and measurement-specific approaches to mediation analysis.