How deeply does Causal Inference for Data Science cover Simpson's paradox?

Explains reversals through stratified examples, confounding, causal graphs, adjustment, and regression, without developing the adjustment formula fully.

Simpson's paradox

As explained in Causal Inference for Data Science

Simpson’s paradox is a statistical phenomenon in which an association or trend observed across aggregated data reverses or differs after conditioning on a grouping variable.

Simpson’s paradox occurs when an association or trend changes, disappears, or reverses after data is grouped by another variable. It warns against treating correlation as causation: a confounder can make an overall comparison point in the wrong direction. section 2.0 section 2.1

Kidney stone treatments

Overall, treatment B appears better, with an 81% recovery rate versus 78% for A. After separating patients by stone size, A performs better for both small stones (93% versus 87%) and large stones (73% versus 63%). Stone size changes the conclusion because treatment groups contain different mixes of small and large stones. section 2.1

Table 2.1 Kidney stone dataset

Treatment	Size	Patients	Recovered
A	Small	87	81
B	Small	270	234
A	Large	263	192
B	Large	80	50

Inspect the kidney stone counts and compare the unstratified recovery table with the size-specific rates to see how aggregation reverses the apparent treatment ranking. section 2.1

Use the paradox as a practical check before making a causal decision:

Ask which variables affect both the choice or treatment and the outcome; these are potential confounders. section 2.0
Compare results within relevant groups instead of relying only on aggregated averages or correlations. section 2.7
Use domain knowledge and causal graphs to decide which variables belong in an adjustment analysis. section 2.9
In regression, include the confounder when appropriate: the unadjusted price model suggests a positive customer trend, while including distance estimates the intended effect as about −1.99. section 6.1.1

How Causal Inference for Data Science covers it

Cover of Causal Inference for Data Science

liveBook manning.com

Read Causal Inference for Data Science on liveBook Get the book at manning.com

Also explained in

liveBook manning.com

Statistics Slam DunkSimpson's paradox is a statistical phenomenon in which an association observed in aggregated data reverses or conflicts with the associations found in relevant subpopulations.

Compare how 2 books explain Simpson's paradox →

Related concepts

correlationcontrasts with · Correlation is a statistical relationship describing how two or more variables or quantities vary together, including the strength and direction of their association, without by itself implying causation.
graphused by · A graph is a mathematical structure consisting of vertices connected by edges that represent relationships or connections.
causal adjustmentcontrasts with · Causal adjustment estimates a causal effect by conditioning on variables that block noncausal pathways and averaging over their target-population distribution.
subpopulationcontrasts with · A defined subset of a broader population that shares specified characteristics or conditions.
confounderaddresses · A confounder is a variable associated with both an explanatory or treatment variable and an outcome, whose influence can distort their apparent relationship or estimated causal effect.

This explainer was generated by Manning's knowledge graph from the text of Causal Inference for Data Science. The section links point to the parts of the book it draws on.