The Subtle Art of Lying with Statistics
summary
Shows concrete ways numbers mislead: using arithmetic versus geometric means (reporting 1.5 hours vs 0.53 hours), hiding deviation (Death Valley mean 61°F while range is 15–104°F), sampling and selection bias (claiming 100% tea drinkers from two people), and deceptive visuals like graph cutouts and pictograms. Gives real examples: Wii 50M owners -> 6% = 3M versus Xbox 30M -> 3.3M, and an 82% headline that actually represents ~2% of respondents. Mentions Prometheus/SLO misuse and warns about p-hacking and confounding factors.
AI-written summary · about
chapters (8)
- 0:00Introduction and Icebreaker
- 0:53About the Speaker and Nginx
- 1:32The Power and Pitfalls of Statistics
- 3:56Understanding Biases in Data
- 13:30Misleading Visualizations
- 18:25Correlation vs. Causation
- 20:31Data Dredging and Manipulation
- 24:33Conclusion and Q&A
related talks
32mThe Subtle Art of Lying with Statistics
43mThe Subtle Science of Misleading with Statistics
40mHow to Lie With Stats and Charts
32mIntroducing Reality into Data Metrics
37mHow to Start Critical Thinking in Data Science
05mPerception and bias and metrics, oh my!
52mAB-tests, lies, damned lies, and statistics.
20mDon't Look. See! Are We Blinded by Data (Visualization)?