update.eoar: update.eoar - Update method for EoAR objects
In tmcd82070/EoAR: Evidence of Absence Regression (EoAR)

Description Usage Arguments Value Author(s) See Also Examples

Update (run more iterations of) an EoAR model object output by eoar().

1 2	## S3 method for class 'eoar' update(obj, niters, nburns = 0, nthins = 1, add = FALSE, quiet = FALSE)

`obj`	An object of class `eoar`. See function `eoar`.
`niters`	Number of additional iterations to do.
`nburns`	Number of additional iterations to burn.
`nthins`	Thinning interval for additional iterations
`add`	If TRUE, the additional `niters/nthin` kept iterations will be appended to the previous iterations. If FALSE, the additional `niters/nthin` kept iterations will overwrite the previous iterations, effectively increasing the burn-in period of the MCMC model.

An eoar model object. See eoar

Trent McDonald

eoar, coef.eoar.

# A 3 year study of 7 sites. 21 "cells". lambda change = 20/year
set.seed(9430834) # fixes Y and g of this example, but not the RNG's used in chains
ns <- 3

ny <- 7
g <- data.frame(
 alpha = rnorm(ns*ny,70,2),
 beta = rnorm(ns*ny,700,25)
)
Y <- rbinom(ns*ny, c(rep(20,ny), rep(40,ny), rep(60,ny)), g$alpha/(g$alpha+g$beta))

df <- data.frame(year=factor(c(rep("2015",ny),rep("2016",ny),rep("2017",ny))),
   Year=c(rep(1,ny),rep(2,ny),rep(3,ny)))

# Uninformed eoar (use low number of iterations because it's and example)
eoa.1 <- eoar(Y~year, g, df, nburn = 1000, niters= 50*10, nthins = 10 )

# Could do this if eoa.1 did not converge
#  1) Consider everything in eoa.1 to be burn-in. (nadapt+nburn+niters iterations)
#  2) Burn another 1000 iterations, then sample 1000 taking every 10th
eoa.2 <- update(eoa.1, nburns=1000, niters=1000, nthin=10)

# Could do the following if eoa.1 converged, but autocorrelation was high.
#  1) Consider everything in eoa.1 to be burn-in. (nadapt+nburn+niters iterations)
#  2) Do another 1000 iterations with higher thinning
eoa.2 <- update(eoa.1, nburns=0, niters=1000, nthin=50)

# Could do the following if eoa.1 converged and autocorrelation were low,
# but simulation error was high (i.e., need more iterations)
#  1) Consider the n.iters in eoa.1 "good".
#  2) Add another 1000 iterations with same (or different) thinning
eoa.2 <- update(eoa.1, nburns=0, niters=1000, nthin=50, add=TRUE)