A visual explanation of Simpson's Paradox

statisticsdata-visualizationcausal-inference

Abstraction: Aggregate trend reverses when data is disaggregated by subgroups

Key points:

Simpson's paradox: overall trend in aggregate data can be opposite to trend within each subgroup
Classic example: newer kidney stone treatment appeared more effective overall, but was preferentially applied to smaller (easier) stones, masking its true inferiority
Texas/Wisconsin school test example: minority students outperform subgroup-by-subgroup in Texas, yet Wisconsin leads state rankings due to demographic composition differences
Interactive visual demonstration by Lewis Lehe and Victor Powell illustrates the mechanism clearly
Practical danger: can lead doctors to prescribe less effective treatments or schools to copy wrong practices

Connections: Simpsons Paradox · Confounding · Statistical Aggregation