wiqid-package: Fast, simple estimation functions for wildlife population...

In wiqid: Quick and Dirty Estimates for Wildlife Populations

Fast, simple estimation functions for wildlife population models

Description

Quick and dirty functions to estimate occupancy, survival, abundance, species richness and diversity, etc. for wildlife populations.

Details

There are a number of sophisticated programs for the analysis of wildlife data, producing estimates of occupancy, survival, abundance, or density. wiqid began as a collection of fast, bare-bones functions which can be run from R suitable for use when you are generating hundreds of simulated data sets. The package takes its name from the quick-and-dirty nature of the original functions.

We now use wiqid in basic wildlife study design and data analysis workshops, and most functions now have options to check the input data and give informative error messages. Workshop participants have used lm, glm and functions in the secr and BEST packages. So wiqid tries to match the look and feel of these functions.

All functions use standard data frames or matrices for data input. ML estimation functions return objects of class wiqid with parameter estimates on the transformed scale (usually logit functions), variance-covariance matrix, and back-transformed ‘real’ values; there are print, logLik and predict methods. Bayesian functions (distinguished by an initial "B") return class mcmcOutput objects.

Simulations and bootstraps often generate weird data sets, eg. capture histories with no captures. These functions do not throw errors or give warnings if the data are weird, but return NAs if estimates cannot be calculated. Errors may still occur if the data are impossible, eg. 6 detections in 5 occasions.

Note that in version 0.2.0 the scaling of continuous covariates has changed to SD=1 (previously SD=0.5). This means that beta coefficients will now be exactly half the size, matching the output from other software.

The functions are listed by topic below.

SIMPLE BAYESIAN POSTERIORS

Bbinomial	generate draws from a conjugate beta posterior distribution
Bpoisson	generate draws from a conjugate gamma posterior distribution
Bnormal	fit a basic normal model to data

OCCUPANCY

Single-season occupancy

occSS	general-purpose ML function; allows site- and survey-specific covariates
BoccSS	general-purpose Bayesian implementation of the above
occSS0	a basic psi(.) p(.) model, faster if this is all you need
BoccSS0	a Bayesian implementation of the psi(.) p(.) model
occSSrn	Royle-Nichols method
occSStime	faster if you have only time effects, also does a plot
occSScovSite	faster if you only have site-specific covariates
occ2sps	single-season two-species models

Multi-season occupancy

occMS	general-purpose function; parameters depend on covariates; slow
occMScovSite	smaller range of covariate options
occMS0	a simple multi-season model with four parameters; faster
occMStime	parameters vary by season; faster

DENSITY from spatial capture-recapture data

We use the secr package for ML estimation of density. For Bayesian estimation, wiqid offers:

Bsecr0

a Bayesian implementation of the intercept-only model

ABUNDANCE from closed-population capture-recapture data

Although data for genuinely closed populations are rare, this is an important conceptual stepping-stone from CJS models to robust models for survival.

closedCapM0	simple model with constant capture probability
closedCapMb	permanent behavioural response to first capture
closedCapMt	capture probability varies with time
closedCapMtcov	allows for time-varying covariates
closedCapMh2	heterogeneity with 2-mixture model
closedCapMhJK	jackknife estimator for heterogeneity

SURVIVAL from capture-recapture data

Cormack-Jolly-Seber models

survCJS	model with time-varying covariates
BsurvCJS	a Bayesian implementation of the above
survCJSaj	allows for different survival for adults and juveniles

Pollock's robust design

survRDah	2-stage estimation of survival and recruitment
survRD	single stage maximum likelihood estimation

Note that the RD functions are preliminary attempts at coding these models and have not been fully tested or benchmarked.

SPECIES RICHNESS from species x sample matrices

Rarefaction

richRarefy	Mao's tau estimator for rarefaction
richCurve	a shell for plug-in estimators, for example...
richSobs	the number of species observed
richSingle	the number of singletons observed
richDouble	the number of doubletons observed
richUnique	the number of uniques observed
richDuplicate	the number of duplicates observed

Coverage estimators

richACE	Chao's Abundance-based Coverage Estimator
richICE	Chao's Incidence-based Coverage Estimator
richChao1	Chao1 estimator
richChao2	Chao2 estimator
richJack1	first-order jackknife estimator
richJack2	second-order jackknife estimator
richJackA1	abundance-based first-order jackknife estimator
richJackA2	abundance-based second-order jackknife estimator
richBoot	bootstrap estimator
richMM	Michaelis-Menten estimator
richRenLau	Rennolls and Laumonier's estimator

BIODIVERSITY INDICES

Alpha diversity

All of these functions express diversity as the number of common species in the assemblage.

biodSimpson	inverse of Simpson's index of dominance
biodShannon	exponential form of Shannon's entropy
biodBerger	inverse of Berger and Parker's index of dominance
biodBrillouin	exponential form of Brillouin's index

Beta diversity / distance

All of these functions produce distance measures (not similarity) on a scale of 0 to 1. The function distShell provides a wrapper to produce a matrix of distance measures across a number of sites.

distBrayCurtis	complement of Bray-Curtis index, aka 'quantitative Sorensen'
distChaoJaccCorr	complement of Chao's Jaccard corrected index
distChaoJaccNaive	complement of Chao's Jaccard naive index
distChaoSorCorr	complement of Chao's Sorensen corrected index
distChaoSorNaive	complement of Chao's Sorensen naive index
distChord	distance between points on a normalised sphere
distJaccard	complement of Jaccard's index of similarity
distMorisitaHorn	complement of the Morisita-Horn index of similarity
distOchiai	complement of the Ochiai coefficient of similarity
distPreston	Preston's coefficient of faunal dissimilarity
distRogersTanimoto	complement of the Rogers and Tanimoto's coefficient of similarity
distSimRatio	complement of the similarity ratio
distSorensen	complement of the Sorensen or Dice index of similarity
distWhittaker	Whittaker's index of association

DATA SETS

dippers	Capture-recapture data for European dippers
distTestData	artificial data set for distance measures
GrandSkinks	multi-season occupancy data
KanhaTigers	camera-trap data for tigers
KillarneyBirds	abundance of birds in Irish woodlands
MeadowVoles	mark-recapture data from a robust design study
railSims	simulated detection/non-detection data for two species of rails
salamanders	detection/non-detection data for salamanders
seedbank	number of seeds germinating from samples of soil
Temburong	counts of tree species in a 1ha plot in Brunei
TemburongBA	basal area of tree species in a 1ha plot in Brunei
weta	detection/non-detection data and covariates for weta

DISTRIBUTIONS

These are convenience wrappers for the related d/p/q/r functions in the stats package which allow for parameterisation with mean and sd or (for Beta) mode and concentration.

dbeta2 etc	Beta distribution with mean and sd
dbeta3 etc	Beta distribution with mode and concentration
dgamma2 etc	Gamma distribution with mean and sd
dt2 etc	t-distribution with location, scale and df parameters
dt3 etc	t-distribution with mean, sd and df parameters

UTILITY FUNCTIONS

AICc	AIC with small-sample correction
AICtable	tabulate AIC for several models
allCombinations	model formulae for combinations of covariates
standardize	a simple alternative to scale.

Author(s)

Mike Meredith

wiqid documentation built on Nov. 18, 2022, 1:07 a.m.