test.LR: Test of the link between sample attributes and matrix of...

Description Usage Arguments Details Value Examples

View source: R/test.LR.r

Description

The function testing the link between sample attributes and species composition of the matrix, from which weighted means are calculated. The test is based either on db-RDA (distance-based redundancy analysis), or Moran's I. Significant relationship is considered as an argument to use modified permutation test instead of the standard permutation test for testing the relationship between weighted mean of species attributes and sample attributes.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
test.LR(M, env, type = "cor", cor.coef = c("pearson", "kendall",
  "spearman"), exact = FALSE, alpha = 0.001, sqrt = F, perm = 499)

test.LR.0(M, env, type = "cor", cor.coef = c("pearson"), exact = FALSE,
  alpha = 0.001, sqrt = F, perm = 199)

## S3 method for class 'testLR'
print(x, digits = 3, ...)

## S3 method for class 'testLR'
summary(object, ...)

## S3 method for class 'testLR'
coef(object, ...)

Arguments

M

Object of the 'wm' class. Matrix with weighted means of species attributes.

env

Matrix with environmental variables.

type

Currently not implemented. In the future, other types of the test (apart to the one based on db-RDA) should be available.

cor.coef

If type = "cor": which correlation coefficient should be used? Partial match to c("pearson", "spearman", "kendal").

exact

If type = "cor" and cor.coef = "spearman" or cor.coef = 'kendal' - a logical indicating whether and exact p-value should be computed.

alpha

Target Type I error rate for Monte Carlo permutation tests (influences number of permutations).

sqrt

Logical value, default FALSE. Should the distance matrix based on Whittaker's index of association be square-rooted to become Euclidean? See Details.

perm

Number of permutations for type = "cor". Default = 199.

x, object

Object of the class "testLR"

digits

Number of digits reported for parameters in summary output.

...

Other arguments passed into print, summary or coef functions (not implemented yet).

Details

In case of dbRDA, the matrix of intersample distances is calculated using Whittaker's index of association (ia) and significance of the variation explained by sample attributes (R2) is tested by Monte Carlo permutation test. In case of Moran's I, the test is examining wheather the sample attributes variable is compositionally autocorrelated, i.e. whether the Moran's I calculated on this variable using as weighted inverted dissimilarities between sample's species composition is significant.

Whittaker's index of association (calculated as Manhattan type distance on species profiles) is metric, but not Euclidean, and in PCoA (on which dbRDA is based) it can produce negative eigenvalues. After square root transformation, the index becomes both metric and Euclidean.

Variation explained by given environmental variable (R2) can differ between individual weighted means calculated from the same species composition matrix. This happens when different species attributes have different values missing; explained variation is calculated only from those columns (species) of compositional matrix L, which have assigned value for given species attribute.

Value

Object of the class "testLR", list of lists, each node containing two parts: results of dbRDA analysis (calculated by capscale function from vegan) and results of Monte Carlo permutation test (calculated by anova.cca function, also from vegan).

Examples

1
2
data (vltava)
test.LR (M = wm (vltava$spe, vltava$ell), vltava$env[,'pH', drop = FALSE], type = 'cor')

zdealveindy/weimea documentation built on Dec. 5, 2017, 11:25 p.m.