test.partitions: Run a set of significance tests on vector plot components.

In ctkremer/priceTools: Tools for analyzing biodiversity-ecosystem function relationships

Description

This functions tests the significance of the changes in richness and in function associated with the vectors of a vector plot. It compares a single baseline (or control) community against a comparison (treatment) community. This can be calculated for the BEF, CAFE, and 5-part Price components.

Usage

test.partitions(data, type = "both", treat.var, control, standardize = T,
  print = F, plot = F)

Arguments

data	Input data, generated by pairwise.price
type	Specify which set of vector components to test ("bef", "cafe", "both", "price")
treat.var	Specify which column in the input data contains the treatment variable
control	Identify which level of the treatment variable shoudl serve as the control distribution
standardize	Use standardized variables? (T/F)
print	Print output table?
plot	Display plot?

Details

Currently, this simple suite of tests focus on detecting differences in the mean and variance of distributions using parametric tests. In future versions, this can (and should) be expanded to include a variety of other tests, including non-parametric approaches. This will be important, as these distributions are often skewed and complex.

Value

Table of significance tests for the means and variances of vector components. variable is the richness or function component being tested, trt.mean and ctrl.mean are the means of each variable for the treatment and control distributions, delta.mean is the difference in means, mn.pvals is the p-value associated with a t-test of the difference between means, delta.var is the difference in variance between distributions, and var.pvals is the significance of this difference.

Examples

# Data setup
data(cedarcreek)
head(cedarcreek)

#Identify one grouping columns
cc2<-group_by(cedarcreek,NTrt,NAdd,Plot)

# Perform pairwise comparisons of all communities in cms identified by comm.id
# (takes ~30 sec)
pp<-pairwise.price(cc2,species='Species',func='Biomass')

# Organize/format the results, and pull out a subset using NAdd.x=0 as the control/baseline site
pp<-group.columns(pp,gps=c('NTrt','NAdd'))

pp1<-pp[pp$NAdd %in% c('0 0','0 27.2'),]
pp1$NAdd<-factor(as.character(pp1$NAdd),levels=c('0 0','0 27.2'))

# Test differences between the BEF vectors of control-control comparisons & control-treatment comparisons
test.partitions(pp1,type='bef',treat.var = 'NAdd',control = '0 0',print=F,plot=F)

# Combine these tests with a visual summary of the different distributions being tested:
test.partitions(pp1,type='bef',treat.var = 'NAdd',control = '0 0',print=F,plot=T)

# These tests consider the x-axis (richness) and y-axis (function) values of all of the vectors underlying the vector plots produced by \code{leap.zig}. For example, these two are compatible:
test.partitions(pp1,type='cafe',treat.var = 'NAdd',control = '0 0',print=F,plot=T,standardize=F)
leap.zig(pp1,type='cafe',group.vars='NAdd',raw.points=F,ylim=c(0,1500),standardize=F)
# NOTE - for these comparisions to be valid, the decision to standardize (or not) must be consisted between test.partitions() and leap.zig()

# This approach can also be applied to the CAFE, Price, or CAFE & BEF vector arrangements:
test.partitions(pp1,type='cafe',treat.var = 'NAdd',control = '0 0',print=F,plot=F)
test.partitions(pp1,type='price',treat.var = 'NAdd',control = '0 0',print=F,plot=F)
test.partitions(pp1,type='both',treat.var = 'NAdd',control = '0 0',print=F,plot=F)

ctkremer/priceTools documentation built on May 28, 2019, 7:49 p.m.