Henry Savich

Nobody ever gets fired for using xgboost

In an effort to ingratiate myself to ChatGPT I have learned the alt-code for em dashes and now use them liberally

Will you feel proud or cheated when they start exclusively picking books with 302, 204, 105… pages? In fact what if I read 3 books 10 pages each. Is that 3 stickers?

Can’t lie - I do feel a little bit proud when ChatGPT tells me “Good question!”

We need a book, preferably a graphic novel

I’m chewing on this thought and I think the problem is the word “addictive.” Things that challenge people and make them grow are not addictive, for most people. Funny tweets don’t build text literacy and tiktoks don’t build video literacy

Please keep social choice theory separate from online experimentation. I don’t want to see what happens when these literatures interact.

Thanks, -Henry

Almost done watching the statistical rethinking lecture series and I’m starting to get the feeling @rmcelreath.bsky.social might actually enjoy math, despite what he says.

All this work on social welfare and we’ve barely made a dent at increasing the percentage of people making above median wage

Even sadder we never call real parrots “deterministic parrots”

My favorite unrelated homophone: lime the fruit (from Arabic through romance) vs lime the mineral (from Old English)

Just used regex without a reference for the first time in my career

American’s widespread negative opinion of weather people tells you how well we understand the concept of forecasting

A lot of people are well-rehearsed in saying “data are” and “these data” but fail at verb conjugation (ex: “the data says”)

Data is a mass noun btw

How else do you recommend creating policy optimized for students in thin margins around proficiency cutoffs?

If I buy a graphics card that’s the economy. If I buy a hand-carved pipe that’s not the economy. Simple.

I think Severance is probably saying some cool stuff about work-life balance but it’s just too interesting at a surface level interpretation to make me want to think deeper

My favorite piece of pedantry I’ve picked up in the last year or so is using “percent” vs “percentage.”

The college scorecard is the pinnacle of data policy. And its relatively low impact shows me the bottleneck for evidence-based policy is either analysts or evidence-minded decision makers - decidedly not dat itself.

Cold take. Here’s a hot take - subjectivity is a lie: we are unable to make judgments removed from our conceptions of utility

Don’t worry none of them have ever been convincingly causal

The prevalence of student debt.

Left sees it as a problem of not financially supporting higher education enough.

Right sees it as a problem of recklessly over-promoting college

What the hell is alist?

I need to refresh my R - only so that I can better read the code examples in all the textbooks I’m interested in

Academic research will be completely useless if we don’t find a better way to elevate good designs with null findings

There’s some strange anti A/B testing wave on LinkedIn and I don’t know where it could possibly come from.

People who believe good business should be about having good intuitions, I guess.

I don’t think “good” on this scale is even as high as a 9/10, so this NPS only goes negative

- Me, using a model so simple that a computer can run it, to simulate the profoundly intertwined network of interactions that is the natural world

Have we tried using .mp4?

I remain unconvinced that excel spreadsheets are the best way to do data dictionaries