PVR: Phylogenetic Eigenvector Regression

In vanderleidebastiani/daee: Data Analysis for Ecology and Evolution

Description

Phylogenetic Eigenvector Regression (PVR) and eigenvector selection.

Usage

PVR(
  traits,
  dist,
  cumulative = 0.99,
  VMoran = 0.025,
  pMoran = 0.05,
  check = TRUE
)

## S3 method for class 'pvr'
plot(x, trait = 1, ...)

Arguments

traits	Species described by continuous traits, with traits as columns and species as rows.
dist	Phylogenetic distance matrix.
cumulative	Percentage of variation in the phylogenetic distances considered in the analysis. Cumulative percentage must be higher than the cumulative percentage of the first two eigenvalues, and less than 1.
VMoran	Stopping rule based on Moran I value (Absolute value, smaller than the specified value).
pMoran	Stopping rule based on the p-value of Moran I (Greater than the specified). check Logical argument (TRUE or FALSE) that checks whether traits and phylogeny taxa labels match. The sequence of species in the trait data must be the same as that in the phylogenetic distance matrix.
check	Logical argument (TRUE or FALSE) to check if species sequence in the traits data follows the same order as the one in the phylodist matrices (Default checkdata = TRUE).
x	An object of class pvr.
trait	Trait for plot.
...	Other parameters for the respective functions.

Details

This function is based on a non-sequential approach, that uses the combination of eigenvectors that minimizes the residual phylogenetic autocorrelation, measured by Moran I. The method can be used to measure the level of phylogenetic signal in ecological data and to study correlated evolution (Diniz-Filho et al 2011).

The phylogenetic distance matrix is double-centered and submitted to principal coordinates analysis (PCoA). This method generates orthogonal eigenvectors that summarize the phylogenetic structure (Diniz-Filho et al 2008).

The sets of eigenvectors is selected with multiple regression model:

Y = a+Xb+e

where Y is a vector describing trait variation in the set of species, X contains a set of k eigenvectors, a the intercept, b is the vector with regression coefficients estimated and e its residuals, the part of variation in Y that is not explained by the X (Diniz-Filho et al 1998).

The function use an iterative search for the eigenvector that reduces the autocorrelation in the residuals. Primarily, the regression for all eigenvectors is calculated, obtaining the residuas. Then, Moran I for each eigenvector is calculated for the residuals. The function select the eingenvector with the lowest Moran I, and then, as new eigenvectors are added to the model, residuals are updated and autocorrelation is reestimated. The search stops when residual autocorrelation is reduced below threshold Moran I specified and when the statistical significance is reached (Diniz-Filho et al 2011).

Value

values	Eigenvalues, relative eigenvalues and cumulative eigenvalues for the PCoA of distance matrix.
vectors	The principal coordinates with positive eigenvalues.
inf.cumulative	Percentage of the variation in the phylogenetic distances considered in the analysis ( The result should be approximately the specified cumulative value).
n.axis.considered	Number of axes considered.
moran.less.than	Morans I value considered in the stopping rules (Absolute value).
p.moran.greater.than	Stopping rule for the p-value.
PSR.curve.axis.x	Values for the PSR curve (Abscissa).
PSR.curve.axis.y	Values for the PSR curve (Ordinate, for each traits).
minimun.moran	Parameters, number of parameters, observed Moran I, p-value for Moran I, R Squared and p-value for regression model that minimize autocorrelation coefficients in the residuals for each trait.

Note

The parameter in minimun.moran is shown as follows y ~ x[,3] + x[,5], in other words Trait~Axis_3+Axis_5 and so on.

Author(s)

Vanderlei Julio Debastiani <[e-mail]>

References

Diniz-Filho, J. A. F., Santana, C. E. R., Bini, L. M. (1998). An eigenvector method for estimating phylogenetic inertia. Evolution, 52(5), 1247-1262.

Diniz-Filho, J.A.F., Bini, L. M., Rangel, T. F., Morales-Castilla, I., Olalla-Tarraga, M. A., Rodriguez, M. A. & Hawkins, B. A. (2011). On the selection of phylogenetic eigenvectors for ecological analyses. Ecography, 35(3), 239-249.

See Also

PVR.adonis

Examples

# require(SYNCSA)
# data(flona)
# Res <- PVR(flona$traits[,1:4], flona$phylo, VMoran = 0.01)
# Res
# plot(Res, trait = 1)

vanderleidebastiani/daee documentation built on Jan. 22, 2021, 2:41 p.m.