meta: Meta-analysis of movement-model parameters

View source: R/meta.chisq.R

metaR Documentation

Meta-analysis of movement-model parameters

Description

These functions estimate population-level mean parameters from individual movement models and related estimates, including AKDE home-range areas, while taking into account estimation uncertainty.

Usage

meta(x,variable="area",level=0.95,level.UD=0.95,method="MLE",IC="AICc",boot=FALSE,
     error=0.01,debias=TRUE,verbose=FALSE,units=TRUE,plot=TRUE,sort=FALSE,mean=TRUE,
     col="black",...)

Arguments

x

A named list of ctmm movement-model objects, UD objects, or UD summary output, constituting a sampled population, or a named list of such lists, with each constituting a sampled population.

variable

Variable of interest. Can be "area", "diffusion", "speed", "tau position", or "tau velocity".

level

Confidence level for parameter estimates.

level.UD

Coverage level for home-range estimates. E.g., 50% core home range.

method

Statistical estimator used—either maximum likelihood estimation based ("MLE") or approximate ‘best linear unbiased estimator’ ("BLUE")—for comparison purposes.

IC

Information criterion to determine whether or not population variation can be estimated. Can be "AICc", AIC, or "BIC".

boot

Perform a parametric bootstrap for confidence intervals and first-order bias correction if debias=TRUE.

error

Relative error tolerance for parametric bootstrap.

debias

Apply Bessel's inverse-Gaussian correction and various other bias corrections if method="MLE", REML if method="BLUE", and an additional first-order correction if boot=TRUE.

verbose

Return a list of both population and meta-population analyses if TRUE and x is a list of population lists.

units

Convert result to natural units.

plot

Generate a meta-analysis forest plot.

sort

Sort individuals by their point estimates in forest plot.

mean

Include population mean estimate in forest plot.

col

Color(s) for individual labels and error bars.

...

Further arguments passed to plot.

Details

meta employs a custom χ^2-IG hierarchical model to calculate debiased population mean estimates of positive scale parameters, including home-range areas, diffusion rates, mean speeds, and autocorrelation timescales. Population coefficient of variation (CoV) estimates are also provided. More details are provided in Fleming et al (2022).

Value

If x constitutes a sampled population, then meta returns a table with rows corresponding to the population mean and coefficient of variation.

If x constitutes a list of sampled populations, then meta returns confidence intervals on the population mean variable ratios.

Note

The AICc formula is approximated via the Gaussian relation.

Confidence intervals depicted in the forest plot are χ^2 and may differ from the output of summary() in the case of mean speed and timescale parameters with small effective sample sizes.

As mean ratio estimates are debiased, reciprocal estimates can differ slightly.

Author(s)

C. H. Fleming.

References

C. H. Fleming, I. Deznabi, S. Alavi, M. C. Crofoot, B. T. Hirsch, E. P. Medici, M. J. Noonan, R. Kays, W. F. Fagan, D. Sheldon, J. M. Calabrese, “Population-level inference for home-range areas”, Methods in Ecology and Evolution (2022) doi: 10.1111/2041-210X.13815.

See Also

akde, cluster, ctmm.fit.

Examples

# load package and data
library(ctmm)
data(buffalo)

# fit movement models
FITS <- AKDES <- list()
for(i in 1:length(buffalo))
{
  GUESS <- ctmm.guess(buffalo[[i]],interactive=FALSE)
  # use ctmm.select unless you are certain that the selected model is OUF
  FITS[[i]] <- ctmm.fit(buffalo[[i]],GUESS)
}

# calculate AKDES on a consistent grid
AKDES <- akde(buffalo,FITS)

# color to be spatially distinct
COL <- color(AKDES,by='individual')

# plot AKDEs
plot(AKDES,col.DF=COL,col.level=COL,col.grid=NA,level=NA)

# meta-analysis of buffalo home-range areas
meta(AKDES,col=c(COL,'black'),sort=TRUE)

ctmm documentation built on Nov. 4, 2022, 5:06 p.m.