DataSciFact's profile picture. Daily data science tweets from @JohnDCook.

Data Science Fact

@DataSciFact

Daily data science tweets from @JohnDCook.

Conditional independence notation X ⫫ Y johndcook.com/blog/2020/03/2…


The sum of (x_i - x)² over a set of data points x_i is minimized when x is the sample mean.


Ridge regression uses an L2 regularizer. Lasso uses L1. Elastic net uses a convex combination of L2 and L1.


Truth is stranger than fiction: Bayesian statistics comes up in the Epstein files. racket.news/p/the-jeffrey-…


By analogy with zero-knowledge technology, you could think of differential privacy as epsilon-knowledge.


The Apollo mission guidance systems relied on SRIF, square root information filters, a numerically robust variation on Kalman filters.


Skewness is a dimensionless version of the 3rd central moment.


Why natural logarithms are natural for regression. johndcook.com/blog/2015/02/0…


Data Science Fact reposted

Liked math and computers Read tweets about data science instead of doing homework in 2010 Needed job because PhD advisor left Got mad about lack of engineering support for data science work and decided to do it myself

What’s the lore behind choosing your career path ?

SokeyeA's tweet image. What’s the lore behind choosing your career path ?


Data Science Fact reposted
Rothmus's tweet image.

F1 score is the harmonic mean of precision and sensitivity.


'Bayesian analysis with proper priors is not clearly distinct from probability theory.' -- James O. Berger


Empirical Bayes in a nutshell: Estimate priors like a frequentist then carry out a Bayesian analysis.


How the central limit theorem began johndcook.com/blog/2010/01/0…


'Why can you add co- to variance to get covariance, but you can't add it to median? Because the result would be funny.' - Jonathan Berkowitz


Loading...

Something went wrong.


Something went wrong.