It might look normal but this distribution will ruin your stats
summary
Contaminated normals are mixtures that add heavy tails and a few outliers and they cut statistical power dramatically — the demo shows 90% vs 20% power between clean and contaminated groups; classic t tests gain type I error under skewness and lose power under heavy tails. The presenter builds a Deepnote interactive notebook to manipulate skew and tails and releases hypothesize Python library ported from randar Wilcox using trimmed means and bootstrap for robust tests.
AI-written summary · about
related talks
29mIt might look normal but this distribution will ruin your stats
01h 22mDoing frequentist statistics with Scipy
32mThe Inspection Paradox is Everywhere
32m"Data Morph: A Cautionary Tale of Summary Statistics"
01h 15mAB-testing by cluster
02h 23mEffective Data Visualization
28mData Morph: A Cautionary Tale of Summary Statistics
40mHow to Lie With Stats and Charts