Valid auto-models for spatially autocorrelated occupancy and abundance data

TL;DR

This paper demonstrates that correct implementation of auto-models with valid neighborhood weightings resolves previous anomalies, improving the reliability of spatial dependence modeling in ecological occupancy and abundance data.

Contribution

It clarifies the conditions for valid auto-model implementation and provides corrected methods and R code for ecological spatial data analysis.

Findings

Invalid neighborhood weightings cause significant estimation errors.

Correct auto-model implementation resolves anomalies in previous analyses.

Validated methods improve accuracy of spatial dependence estimates in ecology.

Abstract

Auto-logistic and related auto-models, implemented approximately as autocovariate regression, provide simple and direct modelling of spatial dependence. The autologistic model has been widely applied in ecology since Augustin, Mugglestone and Buckland (J. Appl. Ecol., 1996, 33, 339) analysed red deer census data using a hybrid estimation approach, combining maximum pseudo-likelihood estimation with Gibbs sampling of missing data. However Dormann (Ecol. Model., 2007, 207, 234) questioned the validity of auto-logistic regression, giving examples of apparent underestimation of covariate parameters in analysis of simulated "snouter" data. Dormann et al. (Ecography, 2007, 30, 609) extended this analysis to auto-Poisson and auto-normal models, reporting similar anomalies. All the above studies employ neighbourhood weighting schemes inconsistent with conditions (Besag, J. R. Stat. Soc., Ser. B, 1974, 36, 192) required for auto-model validity; furthermore the auto-Poisson analysis fails to exclude cooperative interactions. We show that all "snouter" anomalies are resolved by correct auto-model implementation. Re-analysis of the red deer data shows that invalid neighbourhood weightings generate only small estimation errors for the full dataset, but larger errors occur on geographic subsamples. A substantial fraction of papers applying auto-logistic regression to ecological data use these invalid weightings, which are default options in the widely used "spdep" spatial dependence package for R. Auto-logistic analyses using invalid neighbourhood weightings will be erroneous to an extent that can vary widely. These analyses can easily be corrected by using valid neighbourhood weightings available in "spdep". The hybrid estimation approach for missing data is readily adapted for valid neighbourhood weighting schemes and is implemented here in R for application to sparse presence-absence data.