[This is an excerpt from https://causalpython.io]

10/10

#CausalSky

(Spoiler alert): The old truth that violence inspires violence seems to transpire here as well.

R package with the implementation of discussed methods is available here: https://cran.r-project.org/web/packages/geocausal/

📃 Paper: https://arxiv.org/pdf/2504.03464

9/

And all of this without any additional assumptions on spatial spillover or spatiotemporal carryover effects.

A very powerful and promising framework.

And if you're wondering whether airstrikes reduced or increased insurgent violence in Iraq...

8/

Through a series of transformations, we can calculate the avg treatment effect (ATE) over a set of locations and some time period T

On top of this, the framework allows us to also calculate "pixel"-wise conditional avg treatment effects (CATE), where "pixels" represent small geographical areas.

7/

...detail daily US airstrikes and insurgent violence with a high degree of spatial fidelity (typically, to ten meter precision)"

Methodologically, the proposed approach uses Poisson point process, and assigns treatment value to each location s at each time point t.

6/

The authors use their new powerful framework to examine the causal effects of American airstrikes on insurgent violence during the Iraq War of 2003–2011.

The analysis is based on "two declassified datasets from the US military that...

5/

...a new method for working with spatio-temporal data directly.

The proposed approach preserves granular spatial and temporal attributes of microlevel data, allows to estimate heterogeneous treatment effects incorporating spatial moderators, and allows for spatio-temporal mediation.

4/

This approach not only leads to information loss, but also makes it virtually impossible to precisely assess spatial spillover and temporal carryover effects of the intervention of interest.

To address these issues, Mitsuru Mukaigawara and colleagues (Harvard, Dartmouth) propose...

3/

According to Mitsuru Mukaigawara and colleagues, "nearly all" such studies rely on "standard methods of causal inference, (that) aggregate (...) microlevel data into coarse panel data formats that discard geospatial and temporal information."

2/

We'll start sending this week's issue at 9am PT / 12pm ET / 6pm CET

Join 4k+ top researchers and practitioners (it's FREE): https://causalpython.io

#CausalSky

3/3

- A brand-new paper by Mitsuru Mukaigawara and colleagues on handling spatio-temporal causal inference with arbitrary spillover and carryover effects

- Kim Falk's brilliant tutorial on "Bandits in the Wild"

2/

Register here (for FREE): www.causalscience.org

Can't join? Subscribe to get a notification when the recording is out: causalbanditspodcast.com

**Read the report: genderedinnovations.stanford.edu/case-studies...

6/6

#CausalSky

...whether subgroup-specific effects exist, and if so, whether and to what extent they matter.

Our structured discussion will end with a Q&A session, where you'll be able to ask all of us questions that you think matter most.

5/

And if so, is it also safe for my 96-year-old aunt to take the same dose of vitamin B12 that I take?

Finally, if it's safe, is it beneficial for her at all?

Next week, we'll meet with two great statisticians that you've surely heard about: Stephen Senn and Richard Hahn, to discuss...

4/

...female-typical injury tolerance, biomechanics, spinal alignment, and more.

But...

Does it matter for actual safety?

Are human bodies similar and resilient enough to react similarly to what happens in a car accident despite their gender and idiosyncratic characteristics?

3/

They were just scaled-down versions of male dummies and...

Represented the 5th percentile of females in terms of body size and mass (aka the smallest 5% of women in the general population).

These dummies also did not take into account...

2/

My full conversation with Zhijing Jin on causality, agents, and responsible AI: https://www.youtube.com/watch?v=Frb6eTW2ywk

4/4

#CausalSky