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...
Definitely who you want reviewing your owl papers
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!
I never understand what the bots are up to
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.
Hey, open population SMR works after all. First stab.
github.com/benaug/Spati...
Give us 2.0! (but don't break my code 😬)
Added 2 interior years with no sampling.
Probably not so interesting of a scenario. I think a rotating traps scenario could make better use of the telemetry data. It only improves estimates to the extent that minimum known year range an individual is in the population is extended by telemetry data beyond what is known from SCR.
Messing around with Open SCR + telemetry survival. Here, Method A is live capture to deploy a collar, assumed to stay on for 2 years or until they die (starting simple). Method B is some generic capture recapture method, say hair snares. So hair snare in years 1,3,5, live cap in years 2,4.
"Simply develop a small set of biologically plausible hypotheses to compare and ..." 😂
Leveled up this OpenPopSCR model with density covs in year 1 and RSF activity center relocation--runs faster, mixes better. Ran this in 1.5 hours. M=225, 5 years, 144 traps/year, 1156 habitat cells.
Effective sample sizes:
D.beta1 rsf.beta sigma.move
779.5270 397.6123 368.2466
telemetry informs other parameters, too. More precision on survival improves recruitment estimate, can improve N estimates to extent telemetry data documents that inds were in pop before 1st cap-recap detection of after the last one
Looking for candidate motivating data sets for open population SCR methods. Need multiple years of capture-recapture data with concurrent telemetry collars where you have capture recapture data to explain how they were marked. Spatial IPM where telemetry informs survival.
Still not great mixing, but finding that open pop SCR dispersal scale parameters mix better if you don't keep the activity centers in the model for augmented individuals are not in the population and just propose a new s trajectory when proposing to turn them on
The Chandler Clark Jolly Seber parameterization implies Poisson entries as M goes to infinity, but in practice we want to set M as low as we can get away with. This causes the variance in recruits to be less than Poisson variance and it shrinks through time. Here "slack" is M - N.super.
17 years ago. Today, I have to simulate my own bears.
I’m recruiting a PhD and MSC student for fall 2026 working on the movement ecology and conservation of Mexican spotted owls in SW forests and rocky canyonlands. Exciting partnership with Los Alamos National Laboratory. Great vibrant lab group, high impact research! 🦉
gavinmjones.com/opportunities/
Shit, found it already and not even bedtime. Spent an hour looking earlier.
Anyways, on the hunt tonight
But also wake up to jibberish emails sometimes 😂
If anyone goes through my phone after I die, they'll find a bunch of random pictures of code. I bet they will be baffled. Who would guess that I look for bugs in my code while falling asleep and email myself if I figure it out so I don't forget. When I do some of my best work!
They might even prevent you from making dimension shifts at all in some cases. And the problem would arbitrarily depend on what you set data aug limit to.
Someone get on this! 😂
The z=0 individuals have ACs consistent with current model parameters and values. They are not real individuals and should not count in the likelihood when updating RJMCMC indicators. But they do. Only way around this is to not keep z=0 ACs in the model.
Hypothesis: data aug can screw up RJMCMC for Dcov selection in SCR because of the z=0 individuals that still have likelihoods for their activity centers. You can actually have z=0 individuals not have ACs and simulate them when proposing to turn them on. I bet you'd get different inclusion probs.
Anyways, the double data augmentation + group site visitation process with ind detection|site visit gives you a means to model cohesion
I bet the effect of ignoring cohesion is reduced when you don't know individual IDs so SMR may be more robust than SCR in this respect
