iiinova — where your digital self keeps evolving

Cannon "like boom" Cloud @cannoncloud.bsky.social · 23/09/2026

Hotter Cannon for ya. Being replaced with the undergrads in solidarity

110

Cannon "like boom" Cloud @cannoncloud.bsky.social · 03/08/2026

Your scientists were so preoccupied with whether they could, they didn't stop to think if they should

An overhead satellite view from Google Maps showing a highly complex, multi-level highway interchange with a dizzying tangle of overlapping roads, ramps, and overpasses, creating a chaotic "spaghetti junction" look labeled "Action Charenton-le-Pont".

111

Cannon "like boom" Cloud @cannoncloud.bsky.social · 10/04/2026

Friends also don't let friends claim causality from a spatial random forest just by throwing variables at it. I never thought I'd be here, yet here I am 😅. To test for confounding, I compared PDPs and mapped local marginal effects against a municipality fixed-effects spatial sponge. Maybe it works?

A grid of five line charts titled "Stability of Marginal Effects: Biophysical Predictors," comparing a main spatially paired model (solid red line) against a municipality fixed-effects specification (dashed blue line). The charts show Partial Dependence Plots for Distance to forest, Forest density, Soil pH, Soil bulk density, and Fire frequency. Across all five charts, the red and blue lines closely overlap in shape and slope, demonstrating that the estimated marginal effects (the curves) remain highly stable even when accounting for regional unobservables.

A two-panel figure titled "Forest density (+1 pp): Absolute divergence in predicted uplift." The left panel is a map of Brazil showing the difference in predicted probability between a standard model and a fixed-effects model. Most of the map is white, indicating zero or near-zero divergence. The right panel is a histogram showing the distribution of this divergence across all pixels. The distribution is tightly peaked at zero, with a subtitle noting a median divergence of +0.00 percentage points and a 95th percentile absolute divergence of only 1.47 percentage points.

220

Cannon "like boom" Cloud @cannoncloud.bsky.social · 06/01/2026

Not that many blocks. Did you say something mean to Jimmy Kimmel?

210

Cannon "like boom" Cloud @cannoncloud.bsky.social · 12/12/2025

I updated my DiD Visualizer tool for TWFE/OLS. 1. I added a break the model button to correlate treatment effects with treatment timing, making negative weight more likely. 2 Made the heterogeneous treatment effects more responsive to make things wacky with uniform distribution ranges instead 1/2

Graph with outcome on y axis and time on X axis, showing the estimates of TWFE/OLS, BJS and underlying true treatment effect. TWFE/OLS is quite wrong while BJS is accurate.

142

Cannon "like boom" Cloud @cannoncloud.bsky.social · 06/11/2025

Unless they think moving to a oligarchic type system (or something similar) favors incumbents that can use current resources to maintain advantages and extract more rent. Never works out long term when the fascist state eats all, but

tobias saying it might work for us meme.

240

Cannon "like boom" Cloud @cannoncloud.bsky.social · 12/08/2025

I made fun of his style, in case that's any better.

130

Cannon "like boom" Cloud @cannoncloud.bsky.social · 07/07/2025

In your original example you used Vermont as the "bluest, yet still purple" state. I just pointed out in the image Hawaii was more blue. As an aside, the color used in that image is #5E35FE which seems a bit blue, whatever that means. Closest named color is "MediumSlate Blue"

Screenshot showing #5E35FE is kinda "blue" whatever that mean

210

Cannon "like boom" Cloud @cannoncloud.bsky.social · 06/07/2025

In the example you chose to show, quite blue. I'm just pointing to that slight irony in your first image.

Wiki page showing 2008 results for Hawaii with 72% for Obama

010

Cannon "like boom" Cloud @cannoncloud.bsky.social · 09/05/2025

I added to the OLS/TWFE_play_around shiny app the option to include heterogeneous dynamic treatment effects by cohort, and you can click a button to include the BJS estimator. cannoncloud.shinyapps.io/TWFE_OLS_Pla... Share your ugly TWFE plots

030

Cannon "like boom" Cloud @cannoncloud.bsky.social · 07/05/2025

I created a shiny web tool to play around with OL/TWFE so you can teach how this stuff can get so screwy. You can mess with temporal and cohort heterogeneity, treatment timing, whether you have any controls, etc. Share your worst plots! #econsky cannoncloud.shinyapps.io/TWFE_OLS_Pla...

TWFE Event_study plot showing sign reversal

56219

Cannon "like boom" Cloud @cannoncloud.bsky.social · 04/04/2025

I made a helpful meme for all the economist running into various strains of econ-mumbo jumbo out there at the moment.

Meme of tapping the sign saying tariffs shift the supply curve. Intended for comedic effect.

032

Cannon "like boom" Cloud @cannoncloud.bsky.social · 20/03/2025

I was disappointed to hear Georgetown's yellow-bellied response to one if it's students, Badar Khan Suri, and employees illegal detention so I wrote a letter to interim President Groves demanding some backbone. I'm hoping fellow alumnus will as well. I'm happy to share a draft of what I sent.

021

Cannon "like boom" Cloud @cannoncloud.bsky.social · 18/10/2024

Subjective, but -0.5 lets you fill in the 0 and 1 with multiple estimates from different models.

Event study plot with treatment start dash line somewhere between -1 and 0.

000

Cannon "like boom" Cloud @cannoncloud.bsky.social · 18/10/2024

Nice just saw this morning you'll be presenting at out seminar series at Goethe in January. Looking forward to it!

110

Cannon "like boom" Cloud @cannoncloud.bsky.social · 15/10/2024

I used McCloskey a lot in a paper where I reviewed a bunch of big picture books, though I wasn't so convinced by her shtick either. Mainly I came away from that paper thinking the whole theory-of-all-growth books was not such a valuable exercise (besides for the authors).

Break down of authors by primary determinants of the Industrial Revolution

020

Cannon "like boom" Cloud @cannoncloud.bsky.social · 26/09/2024

So I looked at the data from 2000 to 2013 in the UN data, and I also see that decrease in resources, but the trend has completely reversed since then.

Shows UK Domestic Material Consumption from 1995 to 2024. It goes down until 2013 or so and then back up to where it was before.

100