Adam Sacarny
@asacarny.bsky.social
Associate Professor @columbiahpm.bsky.social. Also @nber.org & @j-pal.bsky.social. Economics & health policy. #econsky #healthpolicy #medsky 📈🚞🐈🗽🏳️🌈
http://sacarny.com
Posts represent my views, not my employer's
http://sacarny.com
Posts represent my views, not my employer's
Yes! I think so...
January 13, 2026 at 8:54 PM
Yes! I think so...
Yeah, I feel like the one sided test only makes sense in the world where you know a priori the true coefficient could not be <0. So in this case you'd have to be comfortable saying wow we must have gotten a ton of negative measurement error on beta.
January 13, 2026 at 8:50 PM
Yeah, I feel like the one sided test only makes sense in the world where you know a priori the true coefficient could not be <0. So in this case you'd have to be comfortable saying wow we must have gotten a ton of negative measurement error on beta.
Hmm, I think the idea is that if any of the b_i >>> 0 then the test could reject. But depending on the true values of the b_i it could be low-powered.
January 13, 2026 at 8:47 PM
Hmm, I think the idea is that if any of the b_i >>> 0 then the test could reject. But depending on the true values of the b_i it could be low-powered.
The origin of this is that we prespecified joint one-sided tests in our analysis plan. Then we went to do the analysis and learned that there is no off the shelf tool to run joint one-sided tests, nor is there a well-accepted approach to doing so. Oops!
January 13, 2026 at 7:33 PM
The origin of this is that we prespecified joint one-sided tests in our analysis plan. Then we went to do the analysis and learned that there is no off the shelf tool to run joint one-sided tests, nor is there a well-accepted approach to doing so. Oops!
I think it is testing against the correct alternative, but there is something arbitrary about weighting each outcome equally, and if the "correct" weighting is different then the test could be very low-powered.
January 13, 2026 at 7:32 PM
I think it is testing against the correct alternative, but there is something arbitrary about weighting each outcome equally, and if the "correct" weighting is different then the test could be very low-powered.
I think this is exactly what @scottbarkowski.bsky.social is suggesting
January 13, 2026 at 5:56 PM
I think this is exactly what @scottbarkowski.bsky.social is suggesting
I wish I had the code. But if I remember correctly, we basically did this by running a constrained regression where b1hat=b2hat=b3hat, so you just estimate a single b_hat, and we ran a one-sided test on that estimate. This is the approach with equal weighting on the 3 endpoints.
January 13, 2026 at 5:47 PM
I wish I had the code. But if I remember correctly, we basically did this by running a constrained regression where b1hat=b2hat=b3hat, so you just estimate a single b_hat, and we ran a one-sided test on that estimate. This is the approach with equal weighting on the 3 endpoints.
Ran into a related problem in an RCT. We had 3 coprimary endpoints and wanted to run one-sided tests (b1=b2=b3=0 vs. b1<0 b2<0 b3<0). We ended up using an approach like this one: www.sciencedirect.com/science/arti...
January 13, 2026 at 5:45 PM
Ran into a related problem in an RCT. We had 3 coprimary endpoints and wanted to run one-sided tests (b1=b2=b3=0 vs. b1<0 b2<0 b3<0). We ended up using an approach like this one: www.sciencedirect.com/science/arti...
Once I can get a full keyboard of this I will consider it
December 4, 2025 at 1:33 AM
Once I can get a full keyboard of this I will consider it
👆Came here to say this👆
November 30, 2025 at 2:58 AM
👆Came here to say this👆
For sure. When we wrote the peer effects research letter, we couldn’t have a supplement so even the methods went mostly unexplained. (We cited our analysis plan which did describe them… maybe good enough but not ideal)
November 18, 2025 at 6:16 PM
For sure. When we wrote the peer effects research letter, we couldn’t have a supplement so even the methods went mostly unexplained. (We cited our analysis plan which did describe them… maybe good enough but not ideal)
I wish they were more of a thing for econ papers. 600 words prob too short but 1,000 might work. Like I'm v proud of our paper on peer effects w/ @andrewolenski.bsky.social @mlbarnett.bsky.social (jamanetwork.com/journals/jam...). JHE would have been a logical target but writing costs were high.
November 18, 2025 at 5:06 PM
I wish they were more of a thing for econ papers. 600 words prob too short but 1,000 might work. Like I'm v proud of our paper on peer effects w/ @andrewolenski.bsky.social @mlbarnett.bsky.social (jamanetwork.com/journals/jam...). JHE would have been a logical target but writing costs were high.
Love them for getting a simple fact out! Or reporting a study that might otherwise get file-drawered because costs of writing up the whole thing are too high. One thing that bugged me was that JAMA series journals wouldn't accept supplements for them, but that seems to have changed.
November 18, 2025 at 5:03 PM
Love them for getting a simple fact out! Or reporting a study that might otherwise get file-drawered because costs of writing up the whole thing are too high. One thing that bugged me was that JAMA series journals wouldn't accept supplements for them, but that seems to have changed.
HP OfficeJet 3550 regrets to inform you that your document was not selected for printing. Unfortunately, the number of jobs submitted far exceeded the paper in the paper tray.
If you would like to print, we encourage you to submit another job in the future.
If you would like to print, we encourage you to submit another job in the future.
November 3, 2025 at 7:33 PM
HP OfficeJet 3550 regrets to inform you that your document was not selected for printing. Unfortunately, the number of jobs submitted far exceeded the paper in the paper tray.
If you would like to print, we encourage you to submit another job in the future.
If you would like to print, we encourage you to submit another job in the future.
I believe it’s used all the time in epi / public health for binary outcomes! Coefficients have a nice interpretation as relative risk ratios. And often used for estimating vaccine effectiveness, like in the RCT of the Novavax vaccine for Covid: www.nejm.org/doi/full/10....
Safety and Efficacy of NVX-CoV2373 Covid-19 Vaccine | NEJM
Early clinical data from studies of the NVX-CoV2373 vaccine (Novavax), a recombinant
nanoparticle vaccine against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)
that contains the full...
www.nejm.org
October 24, 2025 at 4:05 AM
I believe it’s used all the time in epi / public health for binary outcomes! Coefficients have a nice interpretation as relative risk ratios. And often used for estimating vaccine effectiveness, like in the RCT of the Novavax vaccine for Covid: www.nejm.org/doi/full/10....
This rules!! Just FYI there is a minor typo on the Scheduling page ("Both Intercity and Commuter trains are scheudled every 15 minutes at consistent and predictable intervals.")
October 20, 2025 at 4:48 PM
This rules!! Just FYI there is a minor typo on the Scheduling page ("Both Intercity and Commuter trains are scheudled every 15 minutes at consistent and predictable intervals.")
Just don't run predictive margins!! They are not the same!
October 15, 2025 at 6:53 PM
Just don't run predictive margins!! They are not the same!
Toying around with this yesterday in Stata, clogit, xtpoisson, and ppmlhdfe were pretty similar in speed. But I just had one level of fixed effects (the group id). I would bet if you have additional fixed effects in the regression ppmlhdfe is far superior.
October 15, 2025 at 6:52 PM
Toying around with this yesterday in Stata, clogit, xtpoisson, and ppmlhdfe were pretty similar in speed. But I just had one level of fixed effects (the group id). I would bet if you have additional fixed effects in the regression ppmlhdfe is far superior.
October 15, 2025 at 6:29 PM
@instrumenthull.bsky.social where are we on this?
October 14, 2025 at 5:30 PM
@instrumenthull.bsky.social where are we on this?