A Structured Approach to Data Analysis

Summary

Argues that before fitting any model, analysts must construct a directed acyclic graph that fixes the estimand and adjustment set in advance, illustrated through a worked Baron-Kenny mediation example connecting education, income, and health with modern bootstrap-based extensions in the mediate R package. Develops a project skeleton with separated raw and processed data folders, numbered analysis scripts, and an RStudio project file, then introduces the tidyverse for the analytic workflow. Closes with data coding, entry, and file management practices, program-file editing and verification routines, and exploratory unconditional associations.