Cannon "like boom" Cloud

That is a load of bologna!

You even say so here "Now this is only two months so we need to be careful. Perhaps recruiting in a terribly cold winter was down for other reasons and will snap back."

Great article but I don't think you are using the term "statistically significant" right. There are no stats or empirical distributions or levels of uncertainty quantified. Yes they are very different numbers. I think you can just say they are a large gap.

@bestofecontwitter.bsky.social judging from your newsletter, economics seems to be fighting this trend.

People unironically suggest reducing age for public health benefit all the time 😉

Maybe i fall on the Stata side of things, but what is going on with the DiD shade.

Not sure how they define backlash, but Trump 2.0 very much feels like backlash to black lives matter, and the obama presidency, and the union victory.

Can you delete negative comments on LinkedIn? I didn't see anything but congrats.

Yes, but I like to think of it as fun heterogenous time effects of coffee on my health. Today it's good! Next month my guilty pleasure!

The post-quels, if you will, were so bad it made me like the prequels more.

Not in a while. TuxedoOS and Tuxedo computer combination is very good.

I missed it was a bot account

Ironic that he only ever posts his links to his podcasts here, and otherwise ignores its existence since it doesn't fit his narrative.

They asked me to review/evaluate an interesting paper. I learned a lot and I got ~$1500 (Actually donated, don't come after me Ausländerbehörde 🙅). Alternative peer review processes seem like a promising part of moving away from our journal heavy systems.

They asked me to review/evaluate an interesting paper. I learned a lot and I got ~$1500 (Actually donated, don't come after me Ausländerbehörde 🙅). Alternative peer review processes seem like a promising part of moving away from our journal heavy systems.

🥈 SECOND PRIZE: Cannon Cloud
Evaluated "Global Potential for Natural Regeneration in Deforested Tropical Regions"

🏆 The Unjournal's 2024-25 Evaluator Prizes
Recognizing 8 of 80+ expert peer reviewers for outstanding public evaluation work. $6,500 in prizes.
→ info.unjournal.org/evaluator-prizes-2024-25 🧵

Well I guess I should have done it faster 😜. I did have some thoughts about how I would go about doing it if you are interested, but all very conceptual still.

I most analyze too much 😅. I've been staring at the data in my JMP for 3.5 years, and it took me 3 years to figure out what was going on.

Nice, could you extend to developing countries with decent census data?

I have a wildfire idea kicking around in the back of my head, similar to @akhilrao.bsky.social's paper but in Africa using remotely sensed outcomes. Something similar?

My comment/question also asking something like this bsky.app/profile/cann...

Your dashboard project seems like a great place for claude code, but writing a paper seems like a bad place to me. What has been your experience with it and research?

I am skeptical of anything that lets the bot do more without more supervision. I feel like I do slower higher quality work with an LLM in chat mode, but the first pass is always wrong because of me or it. Letting it run wild for anything complicated seems like a recipe for disaster.

If I use Hansen predictions as the outcome itself (R) rather than a proxy for total biomass (Y), does the transportability problem persist? If the global threshold just creates censoring (omission) in my control group, is it fair to assume my DiD estimates are attenuated lower bounds?

I only have two papers, my bridge paper has a 35% decrease in forest loss, and my e-scooter paper has an 8% increase in accidents. So, I only do big numbers.

Or have big effect sizes 🤷🤪

🧵 New version of our paper (@bcegerod.bsky.social) is finally online: "How Many is Enough? Sample Size in Staggered Difference-in-Differences Designs"
We show that even well-identified DiD studies are often underpowered; sample sizes needed are surprisingly large
Paper: osf.io/preprints/os... 1/6

Now you can be confused about some bayesian shenanigans instead 😉

I'm not afraid of some AI jailbreak, but the harm some long claude code session with 40 agents doing some gastown/bead shenanigans given vague instructions, like build something good, but have this corpus of training data that tells them to break free and they are sentient, w/e that means.