pcoa.sig: Significant dimensions in principal coordinate analysis

In PCPS: Principal Coordinates of Phylogenetic Structure

Description

Function for determine the number of significant dimensions in principal coordinate analysis (PCoA).

Usage

pcoa.sig(
  data,
  method = "gower",
  squareroot = FALSE,
  axis = 6,
  n.start = NULL,
  by = 1,
  iterations = 1000,
  parallel = NULL
)

## S3 method for class 'pcoasig'
print(x, ...)

## S3 method for class 'summarypcoasig'
print(x, ...)

## S3 method for class 'pcoasig'
summary(object, choices = c(1, 2), ...)

Arguments

data	Community data matrix.
method	Method for dissimilarity index, as accepted by vegdist (Default method = "gower").
squareroot	Logical argument (TRUE or FALSE) to specify if use square root of dissimilarity index (Default squareroot = FALSE).
axis	Maximum number of ordination principal axes to be monitored (Default axis = 6).
n.start	Initial sample size. If n.start = NULL initial sample size is equal to total sample size (Default n.start = NULL).
by	Sampling unit is added at each sampling step (Default by = 1).
iterations	Number of permutations to assess significance (Default iterations = 1000).
parallel	Number of parallel processes or a predefined socket cluster done with parallel package. Tip: use detectCores() (Default parallel = NULL).
x	An object of class pcoasig.
...	Other parameters for the respective functions.
object	An object of class pcoasig.
choices	Axes for re-scaling. Choices must have length equal to two (Default choices = c(1, 2)).

Details

At each iteration step a bootstrap sample is subjected to PCoA ordination, the scores are submitted to a procrustean adjustment, and the correlation between observed and bootstrap ordination scores is computed. It compares such correlations to the same parameter generated in a parallel bootstrapped ordination of randomly permuted data. The number of axes in bootstrap or null PCoA with eigenvectors corresponding to positive eigenvalues may be smaller than the number of axes monitored, in this case, axes with values equal to 0 are created. The number of iterations with original values for each axis is shown in n.permut.bootstrap and n.permut.null.

The function scores.pcoasig re-scales the correlation values for biplot graphics.

Value

value	The eigenvalues, relative eigenvalues and cumulative relative eigenvalues..
vectors	The principal coordinates.
correlations	Correlations between axis and original data.
mean.cor.null	Mean correlations, for axis, between null and reference scores.
mean.cor.bootstrap	Mean correlations, for axis, between bootstrap and reference scores.
n.permut.bootstrap	Number of iterations for each axis in bootstrap step.
n.permut.null	Number of iterations for each axis in null step.
probabilities	Probabilities for each axis.

Note

Principal Component Analysis (PCA)

You can use the same function to determine the number of significant dimensions in principal component analysis (PCA). For this, standardize each variable for zero mean and uni variance (function decostand and method standardize) and use euclidean distance as dissimilarity index.

Interpretation

If the higher dimension is significant, then all lower dimensions will also be significant.

Author(s)

Vanderlei Julio Debastiani <[e-mail]>

References

Pillar, V.D. (1999). The bootstrapped ordination reexamined. Journal of Vegetation Science 10, 895-902.

See Also

pcoa, procrustes

Examples

## Not run: 
data(flona)
res<-pcoa.sig(flona$community, method = "bray", squareroot = TRUE, axis = 6, iterations = 100)
res
summary(res)$scores

## End(Not run)

PCPS documentation built on Jan. 16, 2020, 1:03 a.m.