COS 7-8 - Investigating niche and neutral processes in bird community assembly: Regression with spatial eigenvectors can be more informative than regression on distance matrices

Monday, August 8, 2011: 4:00 PM
8, Austin Convention Center
Tracy A. Pinney and Kevin J. Gutzwiller, Department of Biology, Baylor University, Waco, TX
Background/Question/Methods

The relative importance of niche and neutral processes in community assembly has been the focus of much research in ecology. Statistical methods used to test for these processes also continue to be debated. We hypothesized that niche processes were more important than were neutral processes in structuring communities of summer-resident birds in Texas. We also assessed which of two analytical methods was more informative in terms of the strength of associations between predictors and community composition. We wanted to avoid environmental gradients that could bias inferences toward finding that niche processes were more important. We therefore chose a relatively homogeneous study area. We used multiple regression with spatial eigenvectors and multiple regression on distance matrices to assess evidence for niche and neutral processes at 282 sites. For multiple regression with spatial eigenvectors, we used five eigenvectors as spatial variables and five principal components as habitat variables; for multiple regression on distance matrices, we used a geographic distance matrix as the spatial variable and a matrix of five principal components as the habitat variable. We used variance inflation factors to assess multicollinearity among predictors. Standardized regression coefficients and variance partitioning were used to assess the relative importance of predictors.

Results/Conclusions

Analysis of variance inflation factors indicated multicollinearity among predictors was negligible, which enabled us to draw clear inferences about the relative importance of predictor variables. In the multiple regression with spatial eigenvectors, four of the candidate predictors were statistically significant and individually associated with 1% (one spatial variable) to 6-15% (three habitat variables) of the variation in community composition; overall, this model was associated with 31% of the variation. The importance of the habitat variables in these results suggested that niche processes were more important than were neutral processes in structuring the bird communities we studied. Although the spatial predictor in the multiple regression on distance matrices was statistically significant, the habitat predictor was not significant, and the model was associated with < 1% of the variation in community composition, indicating little support for either niche or neutral processes. Neutral processes received little support from either method, perhaps because the vagility of the bird species we studied negated the spatial effects predicted by neutral theory. Our results also suggest that, within a homogeneous region, multiple regression with spatial eigenvectors can be more informative than multiple regression on distance matrices for addressing questions about niche and neutral processes in community assembly. 

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.