Spatial autocorrelation, model selection and hypothesis testing in geographical ecology: Implications for testing metabolic theory in New World amphibians

Authors

  • Fernanda A.S. Cassemiro
  • José Alexandre Felizola Diniz-Filho
  • Thiago Fernando L.V.B. Rangel
  • Luís Maurício Bini

Abstract

In this paper, we stressed that avoiding significance tests under an alternative model selection framework does not mean that spatial autocorrelation no longer matters, since Akaike information criterion (AIC) is sensitive to the presence of spatial autocorrelation. We exemplify our discussion by analysing species richness patterns of American amphibians, in the context of metabolic theory, to understand how the presence of spatial autocorrelation in data affects data analysis under alternative frameworks of hypothesis testing and model selection. In general, temperature was found to be an important predictor of species richness in both frameworks, although particular predictions of metabolic theory were not fully satisfied when taking spatial autocorrelation into account.

Key words: hypothesis testing; spatial autocorrelation; model selection; Akaike information criterion; macroecology; richness gradients; metabolic theory.

Downloads

Issue

Vol. 2 No. 3 (2007): September/December

Section

Articles

License

I grant the journal Neotropical Biology and Conservation the first publication of my article, licensed under Creative Commons Attribution license (which allows sharing of work, recognition of authorship and initial publication in this journal).

I confirm that my article is not being submitted to another publication and has not been published in its entirely on another journal. I take full responsibility for its originality and I will also claim responsibility for charges from claims by third parties concerning the authorship of the article.

I also agree that the manuscript will be submitted according to the journal’s publication rules described above.