From the archives: Have you (like me!) wondered what the ATT means and how it's different from average treatment effects? I use #rstats to explore why we care about (and how to calculate) the ATE, ATT, and ATU #polisky #episky #econsky www.andrewheiss.com/blog/2024/03... - ThreadSky

andrew.heiss.phd • 101 days ago

From the archives: Have you (like me!) wondered what the ATT means and how it's different from average treatment effects? I use #rstats to explore why we care about (and how to calculate) the ATE, ATT, and ATU #polisky #episky #econsky https://www.andrewheiss.com/blog/2024/03/21/demystifying-ate-att-atu/

1 / 4

Comments

jrgant.bsky.social•101 days ago

📌

nicolillo.bsky.social•101 days ago

📌

danielvaisanen.bsky.social•101 days ago

📌

anygirl99.bsky.social•101 days ago

📌

amiesedani.bsky.social•101 days ago

📌

jonathancorbin.bsky.social•101 days ago

📌

guiterasecon.bsky.social•101 days ago

DOIs for your blog posts - awesome.

guiterasecon.bsky.social•101 days ago

And Zotero gets it mostly right!

andrew.heiss.phd•101 days ago

The magic of Rogue Scholar! https://rogue-scholar.org/

heinersalomon.bsky.social•101 days ago

I am just working on a paper using PSM, and I have been looking to visually prepare the difference between treated and control in one graph. Thanks to you (and @lucystats.bsky.social, apparently) I now have a great set of code. Thank you both!

noahgreifer.bsky.social•101 days ago

Much of this can be done using my R package cobalt too, which interfaces with whatever package you’re using to match

heinersalomon.bsky.social•100 days ago

Of course you also made a package for that, hehe. I am already using matchit(), partly in thanks to your great suggestions on matching I found on cross validated (thanks for both!). I am currently using the plot() command after matchit() for visual representation, but will also check out cobalt.

andrew.heiss.phd•101 days ago

For me, it’s always felt weird to use the ATT to find a treatment effect using only treated people. But this neat paper by @noahgreifer.bsky.social and @lizstuart.bsky.social explains why the ATT is actually useful

(And shoutout to @lucystats.bsky.social for the mirrored histogram idea)

jrgant.bsky.social•101 days ago

📌

episconroy.bsky.social•101 days ago

📌

kevinmci.bsky.social•101 days ago

📌

lizstuart.bsky.social•101 days ago

Thanks for the plug! And just to clarify to estimate the ATT data on the comparison group is used. The inference population is just the treated group.

profkarpo.bsky.social•101 days ago

Thanks, Andrew! These are terrific resources for both scholars and students ... and as an added bonus, I always learn a great deal about making nice figures from your code.

Comments

Posting Rules

Reply