Learning Statistics Will Save Your Life
summary
Shows concrete stats you can use right now: a 169-user dataset with 3,320 badges yields mean ≈20, median 2, mode 1 and standard deviation 36 because test accounts skew results, so report dispersion not just averages. Uses Ruby's benchmark/benchmark-ips and the descriptive_statistics gem to compare string vs symbol (≈28% difference, t huge) and quote styles (tiny, inconsistent t≈3.87). Demonstrates power math (3%→4% needs ~4,800–5,300 samples per arm) and advocates Bayesian A/B with regret to allow peeking and midstream switching.
AI-written summary · about
chapters (17)
- 0:00Intro
- 1:37Basic Descriptive Statistics
- 3:01My Background
- 4:18Three Types of Lies
- 7:34Math
- 8:41Example
- 9:59Inferring
- 14:45AB Tests
- 19:31Ancient Statistics
- 21:50Peeking
- 22:02Regret
- 23:00Be Testing
- 25:22Enrico Fermi
- 26:04Fermi Estimation
- 28:35Piano Tuners
- 28:53Pivot
- 30:23Patrons
related talks
32mStatistics for Engineers
40mHow to Lie With Stats and Charts
27mThe Subtle Art of Lying with Statistics
38mKnow your data: The stats behind your alerts
32mThe Subtle Art of Lying with Statistics
43mThe Subtle Science of Misleading with Statistics
48mHuman data interactions: experiments in how humans analyze data
38mRole of Statistics in the era of Big Data