Statistical inference of correlated evolution between macroecological variables using phylogenetic eigenvector regression
Carregando...
Data
2000
Título da Revista
ISSN da Revista
Título de Volume
Editor
Resumo
In this paper, a recently developed method for comparative data analysis, called
phylogenetic eigenvector regression (PVR), was applied to macroecological data of five groups of
mammals and birds from South America. In these data sets, the relationship between geographic
range size and body length was functional or generated a constraint envelope in the bivariate space,
in which minimum geographic range size increased with body length. Using the PVR, eigenvectors
were extracted from the double-centered phylogenetic distance matrix, derived from phylogenies
based on different sources. These eigenvectors were used as predictors in a multiple regression in
which the response variables were body length and geographic range size. Body size usually displayed
significant phylogenetic inertia, measured by the coefficient of determination (R2) of the PVR
regression model. The partial correlation between these two variables, after controlling for
phylogenetic eigenvectors, varied in the different groups. Only for the primate data set, with 50
species, the correlation disappeared after controlling phylogenetic inertia in both variables. For
the owl data set (29 species), the constraint envelope was transformed in a significant functional
relationship after using the PVR. One thousand simulations assuming a Brownian motion pattern of
phenotypic evolution, with a parametric correlation of input equal to zero, permitted to calculate
the true Type I error of the method at 5% as being around 10% for most data sets. This was considered
to be satisfactory in comparison with other methods, specially with the non phylogenetic standard
correlation (TIPS). Power curves of PVR were also estimated for all data sets, using 5000 simulations
with input correlations ranging from 0.20 to 0.95, and indicated a relatively low statistical power
when samples sizes are smaller than 25 species. In general, the PVR method works fine with
macroecological data and the results supported the importance of controlling for phylogenetic
patterns before using ecological or evolutionary mechanisms to explain geographic range size -
body size relationships.
Descrição
Palavras-chave
Macroecological variables, Phylogenetic eigenvector regression
Citação
DINIZ FILHO, José A. F.; COELHO, Alexandre S. G.; SANT'ANA, Carlos E. R de. Statistical inference of correlated evolution between macroecological variables using phylogenetic eigenvector regression. Ecologia Austral, Buenos Aires, v. 10, n. 1, p. 27-36, 2000.