Data Analysis (conbio-data-analysis)
Conservation Biology reviewers are methodologically sophisticated, and the journal expects a
data-availability statement with data and code deposited at acceptance (see
conbio-reporting-and-data-policy). Analyze as if your code will be re-run — because it may be. This
skill covers execution and reporting norms; design decisions live in conbio-study-design.
When to trigger
- Running main and supporting analyses; building the results section
- A reviewer asked for robustness, alternative models, or uncertainty
- Reconciling exploratory vs. confirmatory analyses
- Making the analysis reproducible before deposit
Analysis norms Conservation Biology expects
- Report uncertainty honestly. Confidence/credible intervals, not just stars or p-values; report
the magnitude and conservation meaning of the estimate, not only significance.
- Use the right model for the data. Hierarchical/mixed models for nested data; occupancy and
N-mixture for detection; capture-recapture for survival/abundance; GLMs/GAMs for nonlinearity;
account for spatial autocorrelation and zero-inflation where present.
- Robustness that probes, not decorates. Show specifications that could break the result
(alternative predictors, samples, priors, estimators), and say what you learn.
- Right inference. Cluster/group at the correct level; avoid pseudoreplication in the analysis;
correct for multiple comparisons when testing many implications.
- Confirmatory vs. exploratory. Separate preregistered/confirmatory tests from exploratory ones;
do not mine for a significant interaction and theorize it post hoc.
- Model checking. Report convergence, residual diagnostics, validation/out-of-sample performance
for predictive models; show the result is not an artifact of one modeling choice.
Conservation-specific reporting
- Translate estimates into decision-relevant quantities (extinction risk, population trend,
effect of a management action, area needed) with uncertainty.
- For projections (PVA, SDM, climate), state the assumptions and the range of plausible outcomes —
not a single point forecast.
Reproducibility while you work (not at the end)
- One master script regenerates every table and figure from the (raw or constructed) data.
- Set and report seeds for bootstrap, MCMC, simulation, and any stochastic step.
- Pin software/package versions (
renv.lock, requirements.txt, recorded installs).
- Keep table/figure numbers matched to script outputs.
Anti-patterns
- Stars/p-values with no effect sizes or intervals
- Raw counts analyzed as abundance with detection ignored
- "Robustness" that only reruns near-identical specs to manufacture stability
- p-hacking / HARKing exploratory results into confirmatory claims
- A single point projection presented as certain
- A results section whose numbers the code cannot reproduce
Evidence pass for Conservation Biology
Use this as a second-pass capability check. First lock the species/system threat, conservation decision, and uncertainty relevant to action; then test whether the manuscript addresses conservation-science reviewers who ask whether evidence changes biodiversity, management, or policy action.
- Primary move: Audit unit, comparison, uncertainty, missingness, sensitivity, and reproducibility before making any prose or submission recommendation.
- Decision ledger: return
claim / evidence / blocker / next edit rows so the next pass can patch the manuscript directly.
- Neighbor test: compare against Biological Conservation for applied conservation breadth, Global Change Biology for climate/ecosystem process, Ecology Letters for theory-forward ecology; if the neighboring outlet has the stronger audience claim, recommend re-routing before polishing.
- Submission-ready gate: before final advice, re-open
resources/official-source-map.md for upload-week rules and name the one live-check item that could change the recommendation.
Output format
【Main estimate】magnitude + interval + conservation meaning
【Model】why this model fits the data (detection / hierarchy / spatial)
【Robustness】specs that could break it → what held
【Confirmatory vs exploratory】clearly separated?
【Uncertainty in projections】range stated, not a point?
【Reproducible】master script + seeds + pinned versions? [Y/N]
【Next】conbio-figures-and-tables
Supplementary resources
1---2name: conbio-data-analysis3description: Use when executing and reporting the analysis for a Conservation Biology manuscript so it survives expert, double-blind review — appropriate ecological/statistical models, honest uncertainty, robustness, and reproducibility. Covers detection, hierarchical models, spatial structure, and effect sizes that matter for conservation. Guides analysis norms; it does not fabricate results.4---56# Data Analysis (conbio-data-analysis)78*Conservation Biology* reviewers are methodologically sophisticated, and the journal expects a9**data-availability statement** with data and code deposited at acceptance (see10`conbio-reporting-and-data-policy`). Analyze as if your code will be re-run — because it may be. This11skill covers execution and reporting norms; design decisions live in `conbio-study-design`.1213## When to trigger1415- Running main and supporting analyses; building the results section16- A reviewer asked for robustness, alternative models, or uncertainty17- Reconciling exploratory vs. confirmatory analyses18- Making the analysis reproducible before deposit1920## Analysis norms Conservation Biology expects21221. **Report uncertainty honestly.** Confidence/credible intervals, not just stars or p-values; report23 the **magnitude and conservation meaning** of the estimate, not only significance.242. **Use the right model for the data.** Hierarchical/mixed models for nested data; occupancy and25 N-mixture for detection; capture-recapture for survival/abundance; GLMs/GAMs for nonlinearity;26 account for spatial autocorrelation and zero-inflation where present.273. **Robustness that probes, not decorates.** Show specifications that could *break* the result28 (alternative predictors, samples, priors, estimators), and say what you learn.294. **Right inference.** Cluster/group at the correct level; avoid pseudoreplication in the analysis;30 correct for multiple comparisons when testing many implications.315. **Confirmatory vs. exploratory.** Separate preregistered/confirmatory tests from exploratory ones;32 do not mine for a significant interaction and theorize it post hoc.336. **Model checking.** Report convergence, residual diagnostics, validation/out-of-sample performance34 for predictive models; show the result is not an artifact of one modeling choice.3536## Conservation-specific reporting37- Translate estimates into **decision-relevant quantities** (extinction risk, population trend,38 effect of a management action, area needed) with uncertainty.39- For projections (PVA, SDM, climate), state the assumptions and the range of plausible outcomes —40 not a single point forecast.4142## Reproducibility while you work (not at the end)43- One **master script** regenerates every table and figure from the (raw or constructed) data.44- **Set and report seeds** for bootstrap, MCMC, simulation, and any stochastic step.45- Pin software/package versions (`renv.lock`, `requirements.txt`, recorded installs).46- Keep table/figure numbers matched to script outputs.4748## Anti-patterns4950- Stars/p-values with no effect sizes or intervals51- Raw counts analyzed as abundance with detection ignored52- "Robustness" that only reruns near-identical specs to manufacture stability53- p-hacking / HARKing exploratory results into confirmatory claims54- A single point projection presented as certain55- A results section whose numbers the code cannot reproduce565758## Evidence pass for Conservation Biology5960Use this as a second-pass capability check. First lock the species/system threat, conservation decision, and uncertainty relevant to action; then test whether the manuscript addresses conservation-science reviewers who ask whether evidence changes biodiversity, management, or policy action.6162- **Primary move:** Audit unit, comparison, uncertainty, missingness, sensitivity, and reproducibility before making any prose or submission recommendation.63- **Decision ledger:** return `claim / evidence / blocker / next edit` rows so the next pass can patch the manuscript directly.64- **Neighbor test:** compare against Biological Conservation for applied conservation breadth, Global Change Biology for climate/ecosystem process, Ecology Letters for theory-forward ecology; if the neighboring outlet has the stronger audience claim, recommend re-routing before polishing.65- **Submission-ready gate:** before final advice, re-open `resources/official-source-map.md` for upload-week rules and name the one live-check item that could change the recommendation.6667## Output format6869```70【Main estimate】magnitude + interval + conservation meaning71【Model】why this model fits the data (detection / hierarchy / spatial)72【Robustness】specs that could break it → what held73【Confirmatory vs exploratory】clearly separated?74【Uncertainty in projections】range stated, not a point?75【Reproducible】master script + seeds + pinned versions? [Y/N]76【Next】conbio-figures-and-tables77```7879## Supplementary resources8081- [`../../resources/external_tools.md`](../../resources/external_tools.md) — modeling, inference, and synthesis packages82- [`../../resources/official-source-map.md`](../../resources/official-source-map.md) — data-availability and reproducibility expectations
brycewang-stanford