Don’t Hold Out On Me: Some thoughts on out-of-sample prediction

And some other thoughts on other topics in statistics and information science

It’s been a few weeks on the blog so we have lots of posts to share. I’ll break them up into categories.

Today: statistics and information science

Tomorrow: sociology of science

Wed: political science

Thurs: everything else

• Plotting truth vs. predicted value

• A study is conducted on two groups. When does it make sense to report two separate estimates, and when does it make sense to just report the pooled estimate?

• Don’t Hold Out On Me: Some thoughts on out-of-sample prediction

• AISTATS ’25 Best Paper award—Margossian and Saul on exact recovery of means and correlation in VI

• Measurement error model Stan fitting struggle: The funnel again rears its ugly head

• Generalizing for Sampling and Causal Inference (my talk 3pm today at the University of Maryland)

• “Either a 2% or a 75% chance of rain”

• “The difference between ‘significant’ and ‘not significant’ is not itself statistically significant” . . . in the wild!

• Dumb statistical models, always making people look bad

• Doxastic Logic: a model of beliefs

• What happened to genetic algorithms?

I now prefer the term “information science” to “statistics,” “machine learning,” and “data science.”