Mediation Models

With our multiple linear regression models, we can apply them to evaluate if a variable is a mediator in the causal framework. While far more advanced mediation approaches exist, this one lecture in BIOS 6611 serves to introduce the basic concepts and an application we can achieve with the material covered so far this semester.

In this lecture we discuss the fundamental models of mediation analysis (which may look pretty familiar to those also seen in our "Confounding and Precision Variables in Linear Regression" lecture). Inference in the form of p-values and confidence intervals for the proportion mediated will be defined, with examples demonstrating how we can implement this mediation framework.

A video for the Biostatistical Methods I (BIOS 6611) course in the Department of Biostatistics and Informatics at the University of Colorado-Anschutz Medical Campus taught by Dr. Alex Kaizer. Slides and additional material available at https://www.alexkaizer.com/bios_6611.

Table of Contents:

00:00 - Intro Song
00:18 - Welcome
00:36 - Mediation Analysis
02:12 - Fundamental Models of Mediation Analysis
02:55 - Inference for Mediation
05:56 - Mediation Example
07:25 - Crude Model
08:22 - Adjusted Model
09:09 - Covariate Model
10:04 - Proportion Mediated
10:57 - Proportion Mediated: Test Statistic & p
12:02 - Proportion Mediated: CI & Conclusion
13:23 - Inconsistent Mediation
14:45 - A Few Extra Notes