The “real” contents list is the menu in the sidebar, but we need to keep track of what pages are planned and what are already available.
- Bayes with JAGS – a tutorial for wildlife researchers: Introductory page – Why Bayes? Why JAGS?
- Using this tutorial
- Bayes as a rule for learning: The idea that we learn by updating our beliefs in the light of new evidence (ie, data).
- Probability: Definitions and terminology.
- History of probability concepts
- A simple model with just one parameter: Example: How many orangutan in the park? Conjugacy, comb method and JAGS.
- MCMC and the guts of JAGS
- LINEAR MODELS
- Simple regression with JAGS: Galton’s data on the heights of mothers and daughters.
- Logistic regression: The socks-in-the-box data analysis with JAGS.
- HIERARCHICAL MODELS
- Hierarchical logistic regression: Adding a block effect to the socks-in-the-box analysis.
- OCCUPANCY
- Basic occupancy model: the salamanders data set, no covariates.
- Occupancy modelling with site covariates: Swiss willow tits data with covariates for occupancy. Species distribution map.
- Categorical covariates
- Goodness-of-fit measures
- Model selection with WAIC and LOO; Bayesian lasso.
- Adding survey covariates: covariates for detection probability.
- Multi-year occupancy models: grand skinks data set.
- Two-species occupancy models.
- ABUNDANCE FROM CAPTURE-RECAPTURE DATA
- Data augmentation: Kanha tigers data
- SECR: Example with rectangular state space
- Irregular habitat patches based on rasters
- SCR with multi-catch and single-catch traps
- Covariates for traps and animals.
- Habitat suitability covariates.
- Open SECR models
- ABUNDANCE FROM REPLICATE COUNTS (N-mixture models)
- A basic N-mixture model with simulated data
- N-mixture models with different distributions (Poisson, negative binomial, zero-inflated Poisson)
- COMMUNITY MODELS
- MSOM as implemented by Marc Kéry.
How about 2-species occupancy.
I think it is interesting
Thanks Soy, I’ve add that to the list – though may be a while before it actually appears, as haven’t covered it in the workshop.
Hi Soy, Two species models now included.
The people I’m working with have vast interests with N-mixture model. Perhaps we could add it in as a subtopic under ‘Population Estimation’…🤷🏻
Thanks Bob. We haven’t actually included that in past workshops but it’s an important type of model and we should do so. It would probably need its own section as quite different to SCR.
Hmm, that’s not quite true, we did include it in the Bangkok workshop in 2017 because folks asked for it. I’ll tidy that up and include.
Hi Bob, Basic N-mixture models (abundance from replicate counts) now included.
Although you have provided details of priors in each modelling approach, I was wondering whether it would be wise to insert a section on priors (somewhere in the probability section above). That way the readers would be prepared to face ‘priors’ in the later part of modelling. Just my thought! Great job in putting this together.
There’s lots that can be said about priors (eg, see here), but I don’t think people need to digest all that before starting the orangutan example. Perhaps a short discussion of weak vs strong priors would be good at the start, plus more discussion along those lines for each model. Very weak priors seem to be the tradition in ecology, which I think is a shame: we should be making more use of existing information, at least if the results are to be used for management.
Later in the community model, could we also add MSAM (like Yamaura et al., 2012 and Wearn et al., 2017 models)? Would that baffle the newest modellers? OR maybe we could save for those who want to explore on their own and not put in this tutorial. 🙂
I’m not at all sure how (or indeed if) we address MSOMs or other advanced models. Should we just tell folks to work through Kéry & Royle (2016) chapter 11?
I also have reservations about these more complex models. The published examples worked, but they do involve more and more assumptions and the estimates have wider and wider CrIs. We are trying to wring information out of a data set which may not be there. Most models will run just fine and give you some sort of estimate even with no data at all; the estimates are coming from the priors.