Null expectations and null hypothesis testing for the species abundance distribution

TL;DR

This paper explores null expectations for species abundance distributions using feasible set approaches, providing methods for statistical testing and model comparison in ecology.

Contribution

It introduces a comprehensive framework integrating feasible set theory with ecological inference, including numerical methods for null hypothesis testing and model estimation.

Findings

Derived probability formulas for SAD under three sampling approaches.

Provided asymptotic results for feasible set analysis.

Developed tools for null hypothesis testing and model comparison.

Abstract

The number of elements (N) and types (S) sampled from an ecological system are among the most powerful constraints on observations of abundance, distribution, and diversity. Together, N and S determine sets of possible forms (i.e., feasible sets) for the species abundance distribution (SAD). There are three approaches to the description of the null SAD (= the average feasible SAD). The first approach is based on the random uniform sampling of surjections. I calculate the probability of a given SAD, given N and S, under this approach (Eq. 4). The second approach is based on the random sampling of compositions. I calculate the probability of a given SAD, given N and S, under this approach (Eq. 8). The third approach is based on the random uniform sampling of partitions. I review the approach, which was developed by Locey & White (2013), and provide some asymptotic results useful for ecologists. The center of a feasible set is a null expectation, which should deviate enough from the alternative models before invoking for biological or ecological mechanisms underlying the SAD. Here, I integrate the feasible set approach with the typical framework of inference in ecology (goodness-of-fit, null hypothesis testing, model comparison, null modelling). I describe how to perform numerical simulations to describe expectations under different approaches to the feasible set. I develop objective or fitness functions to allow the estimation of the most likely SAD using numerical optimization. I provide tools to compare null expectations based on the feasible set approach with the observations, in the context of model comparison and null hypothesis testing.