The Data Analyst's Guide to Cause and Effect
Welcome!
Version
Errata
1
Introduction
2
Causal Graphs
2.1
The Fork
2.2
The Pipe
2.3
The Collider
2.4
Post-treatment bias
2.5
Session info
3
G-methods and Marginal Effects
3.1
Inverse probability weighting
3.1.1
Bootstrapping
3.1.2
‘Robust’ standard errors for IPW
3.2
G-computation
3.2.1
Bootstrapping
3.2.2
Bayesian g-computation
3.3
Session info
4
Adventures in G-methods
4.1
Doubly robust estimation
4.1.1
Bootstrapping
4.1.2
More covariates
4.2
Bootstrapped sub-group analysis
4.3
Complex longitudinal exposure-outcome feedback
4.4
Session info
5
Most of Your Data is Almost Always Missing
5.1
Simulating missingness
5.2
Poststratification
5.2.1
Frequentist poststratification
5.2.2
Bayesian poststratification
5.3
Instrumental variable analysis
5.3.1
Preparing Cohen et al. (2015)
5.3.2
Randomized treatment assignment as instrument
5.4
Bayesian instrumental variable analysis
5.5
Session info
6
More Missing Data
6.1
Simple mean imputation vs. multiple imputation
6.2
“Combine then predict” or “predict then combine”?
6.2.1
marginaleffects
6.2.2
emmeans
6.2.3
Comparison
6.3
Bayesian imputation
6.4
Session info
7
Multilevel Modelling and Mundlak’s Legacy
7.1
Multilevel malaria medicine
7.1.1
Naïve model: Simple intercept and slope
7.1.2
Fixed effects
7.1.3
Fixed effects interacting treatment and group
7.1.4
Random intercepts
7.1.5
Random intercepts and slopes
7.1.6
Plot models in a panel
7.1.7
Regularized vs. empirical estimates
7.2
Simulating Mundlak
7.2.1
Naïve model (ignoring group)
7.2.2
Fixed effects
7.2.3
Random intercepts
7.2.4
Mundlak model
7.2.5
Plot effect estimate distributions
7.3
Education and prosociality: Mundlak in action
7.3.1
Raw data distributions
7.3.2
Naïve model
7.3.3
Fixed effects
7.3.4
Random effects
7.3.5
Mundlak model
7.3.6
Plotting effect of education on religiosity
7.4
Marginal effects in a multilevel model
7.4.1
Predicting the observed sites
7.4.2
Frequentist workflow
7.5
Session info
8
Package Citations
Published with bookdown
The Data Analyst’s Guide to Cause and Effect
Chapter 1
Introduction
Placeholder.