test_cwm: Testing the relationship between weighted-mean of species...

In weimea: Weighted Mean Analysis

Description

Function calculating relationship between weighted-mean of species attributes and sample attributes and performing standard (row based), modified (column based), or max test of significance.

Usage

test_cwm(cwm, env, method = c("cor"), wcor = FALSE, wstand = FALSE,
  wreg = NULL, dependence = "cwm ~ env", perm = 499, test = "max",
  parallel = NULL, p.adjust.method = "holm", adjustP = FALSE)

## S3 method for class 'testCWM'
print(x, digits = max(3, getOption("digits") - 3),
  missing.summary = FALSE, adjustP = FALSE, ...)

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

## S3 method for class 'testCWM'
plot(x, alpha = 0.05, line = 2.5, cex.lab = 1.5,
  par.mar = c(4, 4, 0.5, 0.5), box.col = c("blue", "red"), box.lwd = 2,
  ...)

Arguments

cwm	An object of the class cwm.
env	Vector or matrix with variables. See details.
method	Statistical method used to analyse the relationship between cwm (of class cwm) and env (sample attributes); partial match to 'cor', 'lm' and 'aov'.
wcor	Logical; should the correlation be weighted by rowsums of com (extracted from cwm?). Default FALSE.
wstand	Logical; should the variables in correlation be first weighted-standardized? Default FALSE. If wstand = TRUE, both env and traits are weighted standardized prior to calcuation; weights are derived from com matrix (extracted from cwm); env is weighted by rowsums of com, while traits are weighted by colsums of com.
wreg	NULL (default) or numeric vector of weights used in regression (method = 'lm'). If weights are provided, weighted least squares are calculated instead of ordinary least squares.
dependence	Should cwm be dependent variable and env independent ('cwm ~ env'), or opposite? Applicable only for method = 'lm'. Partial match to 'cwm ~ env' and 'env ~ cwm'.
perm	Number of permutations.
test	Vector of character values. Which test should be conducted? Partial match to 'standard' or 'rowbased' for standard (row based) permutation test, 'modified' or 'colbased' for modified (column based) permutation test, 'max' for max test (selecting the higher from rowbased and colbased result), and 'all' including all three tests. See Details.
parallel	NULL (default) or integer number. Number of cores for parallel calculation of modified permutation test. Maximum number of cores should correspond to number of available cores on the processor.
p.adjust.method	A string indicating the method of P-value adjustement, see p.adjust.methods for possible choices.
adjustP	Logical, default FALSE. Should be the P-values adjusted? If adjustP = TRUE, the last column in the results is adjusted by method selected in p.adjust.method.
x, object	object of the class "cwm", generated by function cwm.
digits	number of digits reported by print method on object of "cwm" class (default is 3).
missing.summary	Logical; should be the summary of values missing in env, cwm and traits be printed along to the result of test_cwm analysis? Default is TRUE.
...	Other arguments for print, summary, coef or plot functions (some not implemented yet).
alpha, line, cex.lab, par.mar, box.col, box.lwd	Graphical parameters for plot function.

Details

Currently implemented statistical methods: 'cor'. Plan to implement also: 'lm' and 'aov'. For fourth corner, please use test_fourth.

Argument env can be vector or matrix with one column. Only in the case of linear regression (method = 'lm') it is possible to use matrix with several variables, which will all be used as independent variables in the model. For ANOVA and Kruskal-Wallis test, make sure that 'env' is factor (warning will be returned if this is not the case, but the calculation will be conducted).

Difference between method = 'lm' and 'aov' is in the format of summary tables, returned by summary.cwm function. In case of 'aov', this summary is expressed in the traditional language of ANOVA rather than linear models.

Both method = 'lm' and 'slope' are based on linear regression and calculated by function lm, but differ by test statistic: while 'lm' is using F value and is testing the strength of the regression (measured by r2), 'slope' is using the slope of the regression line (b). This statistic is added here for comparison with the fourth corner method.

Specific issue related to weighted mean is the case of missing species attributes. In current implementation, species with missing species attributes are removed from sample x species matrix prior to permutation of species attributes among species.

Value

Function cwm returns list of the class "cwm" (with print and summary methods), which contains the following components:

• call Call to the function.

• out Matrix with analysis results (coefficients, statistics, P-values).

• miss Matrix with counts of missing values in env, cwm and traits.

• param List with the setting of the function parameters (arguments).

See Also

cwm, snc

Examples

data (vltava)
CWM <- cwm (com = vltava$herbs$spe, traits = vltava$herbs$traits)
re <- test_cwm (cwm = CWM, env = vltava$env[,c('pH', 'COVERE32')])
re
plot (re)

weimea documentation built on May 2, 2019, 4:44 p.m.