Sign in

Ben Augustine

@bencaugustine.bsky.social
217 followers 146 following 293 posts

Ecological Statistics. I came for the charismatic megafauna and stayed for the problematic megamodels. Skeptic. github.com/benaug

PostsRepliesMedia
Reposted by Ben Augustine
Aimee M. Van Tatenhove @full-o-fulmars.bsky.social · 22/09/2026
Two acoustic integration #postdoc positions at @cornellbirds.bsky.social: Statistical integration of eBird and acoustic data academicjobsonline.org/ajo/jobs/32523 Acoustic abundance estimation with eBird data academicjobsonline.org/ajo/jobs/32728 Please share and reach out with any questions!
academicjobsonline.org
Cornell University, Lab of Ornithology
Job #AJO32523, WDR-00061235 Postdoc Associate eBird-acoustic statistical integration, Cornell University, Ithaca NY, Cornell Lab of Ornithology, Lab of Ornithology, Cornell University, Ithaca, New Yor...
014
Ben Augustine @bencaugustine.bsky.social · 22/09/2026
On one hand, I don't like hiding most of the model in a custom distribution, but on the other hand, the model runs 228x faster here, so.....
010
Ben Augustine @bencaugustine.bsky.social · 04/09/2026
Scratch that. More than 35x faster
020
Ben Augustine @bencaugustine.bsky.social · 04/09/2026
Refactored this Royle-Young model for search encounter surveys of individuals. Between this and a joint N/density intercept update, runs 12x faster on an real world data set.
github.com
GitHub - benaug/RoyleYoung: Royle Young models for area or transect searches where you detect individuals (not their sign). Density and resource selection. Data simulators and files to fit models in ...
Royle Young models for area or transect searches where you detect individuals (not their sign). Density and resource selection. Data simulators and files to fit models in nimble. Known ID versions...
120
Ben Augustine @bencaugustine.bsky.social · 19/08/2026
OK, I think this is done cooking for now. Added examples of fitting several other jolly-seber data augmentation models using Gibbs updates for latent states instead of updating them conditionally. Read to of readme and then scroll down to "alternative data augmentation approaches". So fast!
github.com
GitHub - benaug/Jolly-Seber-N-Prior-DA: Jolly-Seber MCMC samplers using N-prior data augmentation
Jolly-Seber MCMC samplers using N-prior data augmentation - benaug/Jolly-Seber-N-Prior-DA
050
Ben Augustine @bencaugustine.bsky.social · 18/08/2026
Poaster to Professor! Seriously though, congrats!
010
Ben Augustine @bencaugustine.bsky.social · 14/08/2026
nevermind, i deleted that one. MCMC is easier on a relative scale. made 3 versions of that with per capita recruitment as a function of expected N
000
Ben Augustine @bencaugustine.bsky.social · 14/08/2026
Yes, but I figured this could get me to per capita recruitment as a function of expected N (prefer realized N), and it did. Takes a while to compile due to dependencies, but runs fast github.com/benaug/Jolly...
110
Ben Augustine @bencaugustine.bsky.social · 13/08/2026
Expected recruitment here is exactly proportional to time. In the Dirichlet formulation, the prior is centered on exact proportionality but allows variation around it depending on the concentration.
110
Ben Augustine @bencaugustine.bsky.social · 13/08/2026
fixed
100
Ben Augustine @bencaugustine.bsky.social · 13/08/2026
crap, forgot to scale phi, too, fixing that.
100
Ben Augustine @bencaugustine.bsky.social · 13/08/2026
How bout this? github.com/benaug/Jolly... Set up to estimate fixed recruit rate with different time intervals, but can put independent priors on expected recruits per occasion.
100
Ben Augustine @bencaugustine.bsky.social · 10/08/2026
I definitely see that it will be important at least with sparse data. But no plans to actually analyze any data with it at this point.
100
Ben Augustine @bencaugustine.bsky.social · 10/08/2026
not here, but in other uses of dirichlet
200
Ben Augustine @bencaugustine.bsky.social · 08/08/2026
Well, I don't think they can, they just will tell you they can! I'm sure you've seen some wild stuff as a teacher!
000
Ben Augustine @bencaugustine.bsky.social · 07/08/2026
Me: Can you evaluate this via MCMC? ChatGPT: yes (thinks for 5 minutes), here are the results! Me: Did you actually run the MCMC? ChatGPT: No, I lied. I made the numbers up.
140
Ben Augustine @bencaugustine.bsky.social · 03/08/2026
also interested to see how hurdle distributions compare where the probability of zero is unrelated to the positive counts.
000
Ben Augustine @bencaugustine.bsky.social · 03/08/2026
I have not! I don't remember why you are modeling the counts given detection, but you could also try ZT Poisson for number of objects in a cluster. Geometric puts most probability mass on 1 and larger values for same expected value, ZT Pois more on middle values.
110
Ben Augustine @bencaugustine.bsky.social · 01/08/2026
not quite, but closer lol. Writing up open population SMR/spatial IPM methods
000
Ben Augustine @bencaugustine.bsky.social · 25/06/2026
Yea, OK, mixing can be fine with conditional version of SMR, too, with abundant telemetry data. Runs faster, so you can run more iterations per unit time. Poor mixing in conditional model more of an issue for abundance, survival, recruitment. Negbin obsmod here.
000
Reposted by Ben Augustine
Sarah B. Bassing, PhD @sbassing.bsky.social · 24/06/2026
Just over one week left to apply for the Biometric Working Group’s student travel grant. If you’re a student presenting at The Wildlife Society annual conference this fall, consider applying! No need to be a BWG member. Application deadline is July 3. Details in the link.
sites.google.com
TWS Biometrics Working Group - Student Travel Grants
The purpose of the student travel grants is to promote and support student interest in biometrics and the Biometrics Working Group by attending The Wildlife Society annual conference. The number of gr...
003
Ben Augustine @bencaugustine.bsky.social · 24/06/2026
But I haven't met an SMR data set that did not have overdispersed counts so the conditional model that allows other obsmods is important
000
Ben Augustine @bencaugustine.bsky.social · 24/06/2026
With abundant telemetry data and patience, you can get acceptable (but still not good) mixing for yearly RSF movement parameters in open population SMR. At least when marginalizing out individual ID. Not sure if it's worth making a version for the conditional model where mixing already is not great
110
Ben Augustine @bencaugustine.bsky.social · 23/06/2026
(besides group behavior 😂)
000
Ben Augustine @bencaugustine.bsky.social · 23/06/2026
What do people think are the least well understood aspects about spatial mark resight? Things people often get wrong? Assumptions that are hard to meet in practice? etc etc
100
Ben Augustine @bencaugustine.bsky.social · 23/06/2026
Data simulators/model fitting code for marginal version of SCR random thinning model here github.com/benaug/Rando...
000
Ben Augustine @bencaugustine.bsky.social · 22/06/2026
This point is it the discussion, though
000
Ben Augustine @bencaugustine.bsky.social · 22/06/2026
Expected value of thinned lam0 is lam0*theta. Coverage wrt correct target is nominal.
100
Ben Augustine @bencaugustine.bsky.social · 22/06/2026
Though I am embarrassed I did not notice the point about poor coverage of lam0 remains in the paper, I should have done a better job reading the final draft. The lam0 estimate using randomly thinned data is lower than truth by definition. That the CI poorly covers truth is not relevant.
100
Ben Augustine @bencaugustine.bsky.social · 22/06/2026
New random thinning SCR paper
besjournals.onlinelibrary.wiley.com
Making the most of hard‐won data: Incorporating commonly discarded information from unidentifiable samples improves precision and cost‐effectiveness in wolf population monitoring
RT-SCR provides a principled and cost-effective way to recover information from samples that are usually discarded, improving both precision and the reliability of parameter estimates without extra f...
131
Ben Augustine @bencaugustine.bsky.social · 12/06/2026
Friday code fuel genghistron.bandcamp.com/album/signal...
000
Ben Augustine @bencaugustine.bsky.social · 05/06/2026
Spatial mark resight is coming out of my ears
010
Ben Augustine @bencaugustine.bsky.social · 29/05/2026
Surely a stop on the Breaking Bad RV Tour
000
Ben Augustine @bencaugustine.bsky.social · 21/05/2026
Pretty much my dream job. In another life.
000
Ben Augustine @bencaugustine.bsky.social · 21/05/2026
This postdoc looks awesome recruit.visma.com/spa/public/a...
recruit.visma.com
210
Ben Augustine @bencaugustine.bsky.social · 18/05/2026
who wrote this? 😂
010
Ben Augustine @bencaugustine.bsky.social · 18/05/2026
Smithsonian Ecological Data Scientist ad reposted. Sounds more strictly like data science now, less quantitative ecology/statistics?
trustcareers.si.edu
Data Scientist, IS-1560-12
Job Opening: Data Scientist, IS-1560-12 at National Zoo and Conservation Biology Institute in Front Royal, VA.
100
Ben Augustine @bencaugustine.bsky.social · 17/05/2026
Cluster SCR model for perfectly aggregated and cohesive groups where you observe individual ID but not group ID. Pretty sure. 😆 github.com/benaug/Group...
040
Ben Augustine @bencaugustine.bsky.social · 09/05/2026
The trick is to sit closer to the computer
120
Ben Augustine @bencaugustine.bsky.social · 06/05/2026
Truth is 40 groups, 34 detected groups (n), 204 individuals, ZTPois group size parameter lambda.P=5
010
Ben Augustine @bencaugustine.bsky.social · 06/05/2026
Dusted off this model for group SCR with unknown group IDs. It is difficult to determine which updates are needed to fully explore the posterior, but currently works well with higher group D and smaller group sizes. But posterior bottlenecks otherwise. Need to see if I can alleviate those.
110
Ben Augustine @bencaugustine.bsky.social · 29/04/2026
should have gone all the way with "murmured" 😆
120
Reposted by Ben Augustine
Philip Patton @ptpatton.bsky.social · 20/04/2026
Job Alert: Smithsonian's National Zoo and Conservation Biology Institute is hiring a Data Scientist with robust quantitative skills, knowledge of best practices with data management, and an ability to collaborate across units. Based in Front Royal, VA. trustcareers.si.edu/postings/c63...
trustcareers.si.edu
Ecological Data Scientist, IS-1560-12
Job Opening: Ecological Data Scientist, IS-1560-12 at National Zoo and Conservation Biology Institute in Front Royal, VA.
043
Ben Augustine @bencaugustine.bsky.social · 16/04/2026
Definitely who you want reviewing your owl papers
000
Ben Augustine @bencaugustine.bsky.social · 11/04/2026
That would be cool, but I highly doubt it. The pattern is once the bots or whoever find one of my repos they get downloaded a lot all at once and then again every time I push a change. My best guess is LLM training, but I don't understand why so many clones!
010
Ben Augustine @bencaugustine.bsky.social · 11/04/2026
I never understand what the bots are up to
100
Ben Augustine @bencaugustine.bsky.social · 10/04/2026
Open population SMR scenario with marking effort only in years 1 and 3 out of 6. 60% of marks stay on until following year, 40% make it to 2nd year (unless animal dies), so marks persist into sighting only years.
010
Ben Augustine @bencaugustine.bsky.social · 05/04/2026
Hey, open population SMR works after all. First stab. github.com/benaug/Spati...
020
Ben Augustine @bencaugustine.bsky.social · 04/04/2026
Saturday code fuel thecallousdaoboys.bandcamp.com/album/gigant...
thecallousdaoboys.bandcamp.com
Gigantic Parasite Tongue, by The Callous Daoboys
1 track album
010
Ben Augustine @bencaugustine.bsky.social · 31/03/2026
Give us 2.0! (but don't break my code 😬)
010