Week 11 — Adult Sex Ratio Estimation from Camera-Trap Data
1 Week Overview
Week 11 — Adult Sex Ratio Estimation: Comparing Naive, Predicted, and Occupancy-Based Approaches
This week we explore three camera-trap–based approaches for estimating adult sex ratio (ASR) in red deer, using:
Wohlfahrt, S., Edelhoff, H., Leitner, H., & Hackländer, K. (2026). Red deer adult sex ratio: comparing three approaches using camera trap data. Journal of Wildlife Management 90:e70139. https://doi.org/10.1002/jwmg.70139
| Session | Content | Duration |
|---|---|---|
| Session A | Student paper presentation + ASR methods deep-dive + active learning | ~75 min |
By the end of this week, you will be able to:
- Define adult sex ratio (ASR) and explain why accurate estimation matters for ungulate management and forest regeneration.
- Compare three approaches to ASR estimation—naive, predicted (GAM), and occupancy-based (Royle–Nichols)—and articulate the assumptions, strengths, and failure modes of each.
- Implement naive ASR calculation with bootstrap confidence intervals in R.
- Fit and evaluate a GAM for sex prediction from camera-trap covariates, including model diagnostics and confusion matrix interpretation.
- Describe Royle–Nichols occupancy modeling and explain how latent abundance estimates inform ASR.
- Critically evaluate when simple methods (naive ASR) suffice versus when advanced approaches are necessary, using monitoring design constraints as criteria.
2 Before Class: Required Preparation
Read: Wohlfahrt et al. (2026) — full paper. Focus on the Methods and Results sections.
Submit on Canvas before class — written responses to questions
Guiding questions as you read:
- What is the key equation for computing ASR? Under what assumption is the naive approach unbiased?
- Why did the authors choose weeks 9–17 as the comparison window? What ecological/statistical assumption does this satisfy?
- The GAM’s prediction precision was ~75%. What does that mean in practical terms — how does imperfect prediction affect the predicted ASR?
- How does the Royle–Nichols model differ from a standard occupancy model? What additional parameter does it estimate?
- Under what conditions did the three approaches agree? When might they diverge?
3 Session A — Paper Discussion & Methods Deep-Dive (~75 min)
Entry Ticket — 0:00–0:05
On your notecard, write brief answers to:
- What is the formula for adult sex ratio (ASR)? Write it without looking at your notes.
- From Week 10: What is the core problem that occupancy models solve? (one sentence)
- If 40% of your camera-trap sightings cannot be sexed, how might that bias your ASR estimate? In which direction?
Key Concepts: ASR and Three Approaches — 0:05–0:15
Adult sex ratio (ASR) is a fundamental demographic parameter:
\[\text{ASR} = \frac{N_{\text{males}}}{N_{\text{males}} + N_{\text{females}}}\]
The paper compares three approaches to estimate ASR from camera-trap data:
| Approach | What it does | Key assumption |
|---|---|---|
| Naive | Use only sex-identified sightings; exclude unknowns | Unknowns are missing completely at random (MCAR) |
| Predicted | Use a GAM to predict sex for unknown sightings, then recompute ASR | GAM correctly classifies sex; prediction errors are balanced |
| Occupancy (RN) | Royle–Nichols model estimates latent abundance by sex, accounting for detection probability | Population closure during sampling window; constant per-individual detection |
Study snapshot: 24 cameras · 20,250 camera-days · 4,110 adult/subadult sightings · 436 unknowns (~8.9%) · Austrian Alps · Jan 2020–May 2022 · Focal weeks 9–17
Student Paper Presentation
Your classmate will present Wohlfahrt et al. (2026). As you listen:
- Note which ASR approach you find most convincing and why
- Prepare one follow-up question or observation
Jigsaw Activity: Three ASR Approaches —
Expert phase (7 min): You’ll be assigned to one method. With your partner(s):
- Write the key equation(s) for your method on the poster/whiteboard
- List 2–3 assumptions your method requires
- Identify 1–2 scenarios where your method would fail or give biased results
- Prepare a 2-minute teach-back
| Group | Method | Key focus |
|---|---|---|
| A | Naive ASR | \(\text{ASR} = N_m / (N_m + N_f)\); when does excluding unknowns bias estimates? |
| B | Predicted ASR (GAM) | How does the GAM predict sex? What happens at the 0.5 threshold? |
| C | Occupancy RN | \(P(\text{detect} \geq 1 \mid N, r) = 1 - (1-r)^N\); what is latent abundance? |
Teach-back phase (7 min): Regroup into mixed teams (one person from each group). Each expert teaches their method in ~2 min. Ask questions!
Choose-Your-Own-Method Scenario — 0:46–0:56
Your group will receive a wildlife monitoring scenario. Decide:
- Which ASR approach(es) you would recommend
- Why that approach is the best fit for the constraints described
- What risks or limitations remain
Write your recommendation on your poster.
Structured Mini-Debate — 0:56–1:06
Motion: “Naive ASR is sufficient for most wildlife management applications.”
- Side A — “Keep It Simple”: Argue for naive ASR
- Side B — “Model It Right”: Argue for occupancy-based or predicted ASR
Each side prepares arguments (2 min), then presents (2 min each), followed by open rebuttal (2 min) and synthesis (2 min).
Concept Map + Muddiest Point — 1:06–1:12
On your notecard:
- Draw a quick concept map linking: ASR, naive, predicted, GAM, occupancy, RN, detection probability, latent abundance, bootstrap CI, camera placement, unknowns. Connect terms with labeled arrows (e.g., “assumes,” “accounts for,” “biased by”).
- Write your muddiest point: the one concept from today that is least clear.
- Discuss with a neighbor — can they clarify?
Minute Paper — 1:12–1:15
On your notecard:
- What is the single most important insight about ASR estimation from today?
- Which of the three approaches would you use in your own research system, and why?
- One question you still have heading into the workshop.