Inside the Advanced Analysis Panel

Overview

generate_advanced_panel() compiles a compact, multi-plot summary of your simulation. It is designed for teaching and quick diagnostics. This vignette explains each component, how it is computed, and what patterns to look for.

Enable via P$ADVANCED_ANALYSIS = TRUE or call the function directly on a results list.

What’s included

Rank–Abundance (SAD) — Observed vs. theoretical Fisher log-series.
Occupancy–Abundance — Total abundance vs. number of occupied sites.
Species–Area (specaccum) — Mean richness and variation vs. sampled area.
Distance–Decay — Sørensen dissimilarity vs. geographic distance.
Rarefaction — Expected richness vs. sample size per site.

(If spatstat tooling is installed in your environment, you can optionally add point-process diagnostics; see the “Extensions” section.)

Example

library(spesim)
P <- load_config(system.file("examples/spesim_init_complete.txt", package = "spesim"))
P$ADVANCED_ANALYSIS <- TRUE
res <- spesim_run(P, write_outputs = FALSE, seed = P$SEED)
generate_advanced_panel(res)

Panels in detail

1) Rank–Abundance Distribution (SAD)

Observed: counts per species from res$species_dist (points in all quadrats).
Theoretical: Fisher’s log-series using FISHER_ALPHA, FISHER_X and dominant fraction (species A).
Interpretation: A good match (on log-scale) suggests your realised abundances follow the intended SAD; large departures imply strong spatial filtering or very small sample sizes.

2) Occupancy–Abundance

Data: site × species matrix (res$abund_matrix).
Metric: TotalAbundance (column sum) vs. Occupancy (non-zero sites).
Interpretation: Many systems show a positive trend: common species occupy more sites. Outliers (high abundance but low occupancy) may be habitat specialists or spatially clumped species.

3) Species–Area (specaccum)

Method: vegan::specaccum(..., method = "random") over sites.
Plot: Mean richness with ±1 SD ribbon.
Interpretation: Curves that rise slowly suggest high redundancy; steep curves imply high heterogeneity/beta diversity.

4) Distance–Decay

Geography: Euclidean distance among quadrat centroids.
Dissimilarity: Sørensen/Bray on presence–absence.
Interpretation: Positive slope → communities become more different with distance; reflects environmental turnover and/or dispersal limitation.

5) Rarefaction

Method: vegan::rarefy() by site (quadrat) across increasing subsample sizes (equivalent to the computation done by vegan::rarecurve(), but without plotting side-effects).
Interpretation: Compare richness at equal sample sizes; plateauing curves suggest adequate sampling.

Extensions (optional)

If you install spatstat.geom (and optionally spatstat.explore), you can extend diagnostics with:

Ripley’s K and L(r)-r transforms
Pair correlation g(r)

These help validate the spatial point process (CSR, clustering, inhibition). Integrate similar plots alongside or beneath the default panel as desired.

Reproducibility notes

All panel components are deterministic functions of res.
The SAR component uses random permutations internally (vegan::specaccum(method = "random")); set set.seed() beforehand for reproducibility if you re-run it.
Rarefaction here is deterministic (via vegan::rarefy()).

spesim team

2026-02-23