Data Science Fact · Apr 30, 2026 · 1:31 AM UTC

Data Science Fact

593Media

Pinned Tweet

Data Science Fact@DataSciFactApr 30

“Statistics is hard, especially when effects are small and variable and measurements are noisy. There are no quick fixes … and a formulaic approach to statistics is a principal cause of the current replication crisis.” — Blakeley McShane

5,749

Data Science Fact · Oct 19, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFact6h

Kleinberg's theorem: There are no clustering algorithms that satisfy scale invariance, richness, and consistency.

2,721

Data Science Fact · Oct 19, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFact9h

'I am the very model of a professor statistical' johndcook.com/blog/2022/10/0…

The Very Model of a Professor Statistical

Statistician George Box's rewriting of a song by Gilbert & Sullivan

johndcook.com

3,468

Data Science Fact · Oct 16, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 16

The Wilcoxon Rank Sum Test can have greater asymptotic relative efficiency than the t-test for heavy-tailed distributions.

5,022

Data Science Fact · Oct 16, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 16

'To calculate is not in itself to analyze.' -- Edgar Allen Poe, Murders in the Rue Morgue

5,299

Data Science Fact · Oct 15, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 15

Earth-mover distance and t-closeness johndcook.com/blog/2018/11/0…

Earth-Mover Distance | Wasserstein metric | t-closeness

t-closeness is a privacy criterion, a way of saying groups are not too different from a global distribution, with distance measured by the "earth mover" metric.

johndcook.com

3,811

Data Science Fact · Oct 15, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 15

Unbiased estimators based on complete sufficient statistics are unique. -- Lehmahn-Scheffé theorem

5,100

Data Science Fact · Oct 14, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 14

Reference comparing R, Python, and MATLAB syntax mathesaurus.sourceforge.net/…

4,468

Data Science Fact · Oct 14, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 14

'A Bayesian is one who, vaguely expecting a horse, and catching a glimpse of a donkey, strongly believes he has seen a mule.' ~ @stephensenn

6,123

Data Science Fact · Oct 13, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 13

What does it mean to call noise white, pink, gray, etc? johndcook.com/blog/2015/12/0…

Noise colors: White, Pink, Red, etc.

What the various colors of noise mean and what inspired their names.

johndcook.com

3,946

Data Science Fact · Oct 13, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 13

If λ(x) is a LRT statistic for a point-null hypothesis, asymptotically -2 log(λ) has a chi-square(1) distribution.

6,105

Data Science Fact · Oct 12, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 12

Apophenia: The tendency to see patterns in random data.

6,108

Data Science Fact · Oct 12, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 12

Testing one factor at a time won't discover interaction effects. johndcook.com/blog/2022/10/3…

A/B testing doesn't find interaction effects

Optimizing parts one factor at a time will not optimize the whole. Interactions of factors matter, and they can't be discovered by testing one factor at a time.

johndcook.com

5,100

Data Science Fact · Oct 11, 2026 · 8:55 PM UTC

Data Science Fact@DataSciFactOct 11

The problem is noise, not sample size. statmodeling.stat.columbia.e…

6,596

Data Science Fact · Oct 9, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 9

'An ounce of replication is worth a ton of inferential statistics.' -- J. H. Steigerus

6,211

Data Science Fact · Oct 9, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 9

'Polynomials are notoriously untrustworthy when extrapolated.' -- William Cochran and Gertrude Cox

6,274

Data Science Fact · Oct 8, 2026 · 6:15 PM UTC

Data Science Fact@DataSciFactOct 8

Avoiding Bayesian underflow johndcook.com/blog/2012/07/2…

Avoiding underflow in Bayesian computations

In integration problems coming from Bayesian statistics, it's easy to run into problems with underflow. Here's how to avoid the problem.

johndcook.com

6,211

Data Science Fact · Oct 8, 2026 · 2:55 PM UTC

Data Science Fact@DataSciFactOct 8

'If we know that our individual errors and fluctuations follow the magic bell-shaped curve exactly, then the resulting estimates are known to have almost all the nice properties that people have been able to think of.' -- John Tukey

6,117