Week 8
Day 1
Today
- Please bring your laptop today
In Class Discussion led by Maddie and Camilla (Link)
Overview of paper’s goals and questions:
How would you describe what the models are actually predicting for each site? And what is it NOT?
What are the different hierarchies in this approach/models?
The authors mentioned a few times that their data was non-random - how come?
What are some potential biases in the PIKE data collection that were mentioned?
Can the data tell us anything - not true random sample, uneven effort, heterogeneous resources? Or does Bayes account for all that uncertainty (as the authors expect)?
Does this adequately address the question with important conservation implications on the line/is it appropriate?
Bonus Questions:
(Fun question) What’s one term/phrase the paper used that Brian has mentioned personally disliking?
What were the main takeaways regarding poaching trends and factors influencing poaching the authors felt comfortable inferring from their model results?
How did they handle baseline differences in countries and sites within countries?
African / Asian data: Should the authors split out the analysis?
In Class: Simulating and fitting a Bayesian GLMM
Assigned
- Readings for next class:
Day 2 (lab)
Today
Brian is absent today.
No lab assigned.
In Class Discussion: Dr. Quresh Latif, research scientist from Bird Conservancy of the Rockies, will be visiting (10am-11:30am) to lead a discussion on his paper Designing count-based studies in a world of hierarchical models.
Some questions to think about:
- How do “hierarchical models place greater demands on data”?
- What exactly is a “sub-model”?
- The authors distinguish between ecological processes and observation processes. Why is it important to model these processes separately, and what problems arise when they are confounded?
- Under what conditions does detectability confound estimation of environmental relationships?
- What advantages and disadvantages are there between data from unmarked and marked populations?
- What influences one’s choice of ‘population metric’? For example, how should researchers choose between abundance, density, occupancy, or occupancy dynamics as the primary population metric for a study?
- How can poor sampling design affect the validity of a hierarchical model, even when the statistical model appropriately accounts for imperfect detection?
Assigned
- Following the discussion, post on the discussion board on canvas.
- Readings for next class: