How to Start Critical Thinking in Data Science
summary
Begins with concrete examples: the birthday paradox (23 people → ~50% match), hospital survival numbers that flip when critical-case counts are revealed (900/1000 vs 800/1000 but 30/100 vs 210/400), pie charts not summing to 100, axes not starting at zero, and nutrition labels using a 0.49 serving size trick. Demonstrates cherry picking with Nicolas Cage vs drownings and the WWII plane survival error, explains multiple testing and p = 0.05 plus the green jelly‑bean false positive from 20 tests, and urges domain knowledge and explicit bias accounting.
AI-written summary · about
chapters (20)
- 0:00Critical thinking intro
- 1:13Large data scales break intuition
- 2:18Misleading stats examples
- 4:27Data generation sources
- 5:33Data processing risks
- 6:36Data viz pitfalls
- 9:25Pie/bar chart contexts
- 11:13Line charts missing context
- 12:37Correlation cherry picking
- 14:50Dentist recommendations variables
- 15:53Funding manipulation risks
- 17:42Biases in data analysis
- 18:23Confirmation bias explained
- 19:05Sampling bias explained
- 20:32Survival bias examples
- 22:01Marketing bias success
- 23:06Availability bias risks
- 24:32Causality vs correlation
- 26:38Multiple testing false positives
- 29:35Q&A
related talks
32mIntroducing Reality into Data Metrics
40mHow to Lie With Stats and Charts
27mThe Subtle Art of Lying with Statistics
21mIs It All About the Data?
48mCommon Data Science Mistakes
14mDesigning a Socially-Critical Data Science Course
20mDon't Look. See! Are We Blinded by Data (Visualization)?
47mA Practical(ish) Introduction to Data Science