CVTest: Constant Volatility Test for WFP

In reptalex/WrightFisher:

Description

Constant Volatility Test for WFP

Usage

CVTest(ntests, R, regress = "f", formula = NULL, varformula = NULL,
  tm = NULL, ...)

Arguments

ntests	Positive integer - number of randomly selected constant-volatility transformations (CVTs) for heteroskedasticity testing
R	Compositional matrix - columns must be species and rows timepoints.
regress	Either "f" or "t" - whether regrssion should be performed on the state of the CVT or on the time.
formula	Formula for regression of the jumps in CVTs vs time or state of cvts. Should be in form "DF ~ f" or "DF ~ tm + I(tm^2)" or etc. Default is quadratic regression
tm	Input vector of timpoints to be used if regress='time'
varformla	Input argument for bptest. Default is quadratic regression.

Value

vector of p-values from Bruesh-Pagan heteroskedasticity tests.

References

http://biorxiv.org/content/early/2016/03/18/044495

Examples

library(plotrix)
library(WrightFisher)

set.seed(2)

#this may take ~20-30 seconds
X <- wfp(nSpecies=15,lambda=25)

t <- X$time
R <- X$Community

par(mfrow=c(1,3))
stackpoly(R,stack=T,main='Community Dynamics',xlab='time',ylab='Abundance')

Pvalsf <- CVTest(100,R)
Pvalst <- CVTest(100,R,regress='time')

plot(ecdf(Pvalsf),xlab='P',ylab='#',main='P-values from regress="f"')
lines(c(0,1),c(0,1),lwd=4,col='blue')

plot(ecdf(Pvalst),xlab='P',ylab='#',main='P-values from regress="time"')
lines(c(0,1),c(0,1),lwd=4,col='blue')

#Uncorrected KS test - the P-value from the KS-test gives a comparable estimate of the incompatibility of the simulation with a WFP
ks <- ks.test(Pvalsf,runif(length(Pvalsf)))
ks$p.value

ks.test(Pvalst,runif(length(Pvalst)))$p.value

reptalex/WrightFisher documentation built on May 27, 2019, 5:54 a.m.