ReCAP_sampler: Integrated Population Model for Harvesting and Aerial...

In YunyiShen/ReDDLeslie: Reconstruct Culling and Aerial count of Population

Integrated Population Model for Harvesting and Aerial Counting data under Bayesian Framework

Description

Main sampler of the posterior distribution of vital rates and harvest count in ReCAP.

Usage

ReCAP_sampler(Harv.data, Aerial.data
, nage, mean.f, mean.s
, mean.SRB, mean.H
, mean.A, mean.b = Harv.data[, 1]
, n.iter = 50000, burn.in = 5000, thin.by = 10
, al.f = 1, be.f = 0.05, al.s = 1, be.s = 0.1
, al.SRB = 1, be.SRB = 0.05
, min.aK0 = list(matrix(-0.001, nage[1], 1), matrix(-0.001, sum(nage), 1), 0)
, max.aK0 = list(matrix(0.001, nage[1], 1), matrix(0.001, sum(nage), 1), 500)
, al.H = 1, be.H = 0.05, al.A = 1, be.A = 0.05
, Assumptions = list()
, start.sigmasq.f = 0.05, start.sigmasq.s = 0.05
, start.sigmasq.SRB = 0.05, start.sigmasq.H = 0.05
, start.sigmasq.A = 0.05
, prop.vars = list()
, proj.periods = (ncol(Harv.data) - 1), estFec = T
, estaK0 = F
, aK0 = list(matrix(0, nage[1], 1), matrix(0, sum(nage), 1), matrix(0, 1, 1))
, global = T, null = T, point.est = mean, verb = FALSE, s.tol = 10^(-10))

Arguments

Harv.data	Harvest data, rows to be age classes and columns for years.
Aerial.data	Aerial count data, should be only one row and same number of columns with harvest.
nage	Number of age classes, female first.
mean.f,mean.s	Best estimation of fecundity/survival. Number of rows and columns should be less than harvest, after multiply by Assumptions matrices (see Assumptions), a full fecundity matrix (i.e. assign fecundity for each age class in each projection year) should have one column less and same number of rows with harvest data(one less year interval). Default input (i.e. no specific Assumptions matrix) should be a full fecundity matrix by itself, i.e.: nrow(mean.f) == nrow(Harv.data) & ncol(mean.f)==ncol(Harv.data)-1. nrow(mean.s) == nrow(Harv.data) & ncol(mean.s)==ncol(Harv.data)-1.
mean.SRB	Similar to mean.f, sex ratio at birth, but note SRB is assumed having no age structure. Thus a full SRB matrix should only have one row.
mean.H	Similar to fecundity, for harvest rate. Difference is since we assume a pre-repoduction harvest, a full harvest rate matrix should have same size of harvest data.
mean.A	Similar to H, Aerial count detection rate. By default, a full aerial count matrix should have same size as aerial count data.
mean.b	Best estimation of baseline harvest, usually just firest year's harvest count.
n.iter	Number of posterior sample saved.
burn.in	Burn in iteration.
thin.by	Sample thined by?
al.f,al.s,al.SRB,al.H,al.A	Hyperparameter for prior distribution of fecundity/survival/SRB/Harvest rate/Aerial Detection Rate's variance. Variance of vital rates v beside aK0 follow inverse Gamma distribution with parameter al.v and be.v, v can be fecundity, survival, harvest rate, aerial count detection rate and SRB.
be.f,be.s,be.SRB,be.H,be.A	Hyperparameter for prior distribution of fecundity/survival/SRB/Harvest rate/Aerial Detection Rate's variance. Variance of vital rates v beside aK0 follow inverse Gamma distribution with parameter al.v and be.v, v can be fecundity, survival, harvest rate, aerial count detection rate and SRB.
min.aK0,max.aK0	Prior distribution of aK0s. If fit density dependent Leslie matrix model: fecundity(Population) = (1 - aK0 [[1]] * (Population - aK0[[3]])) * Fec survival(Population) = (1 - aK0[[2]] * (Population - aK0[[3]])) * Surv In which fecundity is a function of population and Fec and Surv is fecundity and survival at population aK0[[3]] aK0's component were assumed to be uniformly distributed with parameter min.aK0 and max.aK0.
Assumptions	Should be a list, each element is a list and named Fec, Surv, SRB, Harv, AerialDet, aK0 for fecundity, Survival, SRB, Harvest rate, Aerial count detection rate, and linear coeffient of Density Dependency. In each list, there should be two matrices, naming age and time. In sampler function, a full vital rate matrix (i.e. assign vital rates for each age class in each year needed, for repoduction related, will be one year less than harvest and aerial count related viral rates) used in projection will be produced by matrix multiplication, e.g. Fec_full <- Assumptions$Fec$age %*% mean.f %*% Assumptions$Fec$time By default, all Assumptions are identity matrices, i.e. prior mean vital rates matrices should be a full one by itself.
start.sigmasq.f	Starting value of fecundity's variance.
start.sigmasq.s	Starting value of survival's variance.
start.sigmasq.SRB	Starting value of SRB's variance.
start.sigmasq.H	Starting value of harvest rate's variance.
start.sigmasq.A	Starting value of aerial count detection rate's variance.
prop.vars	Proposal variance, should be a list named fert.rate, surv.prop, SRB, A,H, aK0, and baseline.pop.count, same dimension with mean.f, mean.s, mean.SRB, mean.A, mean.H, min.aK0, mean.b.
proj.periods	How many years need to be projected? Usually ncol(Harv.data)-1
estFec	Estimate fecundity? Bool. If not, mean.f will be used.
estaK0	Estimate density dependency parameters? Bool. Note this will be masked if set null=TRUE.
aK0	If not estimate aK0, what aK0 is.
global	Whether aK0[[3]] is global or age specific (i.e. will there be age specific carrying capcacity). Bool.
null	Null model for density dependency? Bool. If null=TRUE, no density dependency will be considered, but model can fit time inhomogeneous vital rates.
point.est	How to do point estimation? Can be mean or median.
verb	Want a verbose run? Bool
s.tol	Tolerant for lowest survival rate (below will be considered as 0).

Details

Do not use Assumptions unless you understand it.

Value

A list with length 7:

mcmc.objs	A list of mcmc objects returned by the sampling algorithm
log.like.mcmc	mcmc object for log likelihood
alg.stats	algorithm statistics, e.g. acceptance rate
model.checking	Model checking, including DIC, absolute difference calculated at point estimation and standard deviation of fitted values
fixed.params	fixed parameters
start.vals	starting value of the algorithm
alg.params	parameters for algorithm setting

Author(s)

Yunyi Shen

Examples

##Use our Chicago Deer data.
set.seed(42)
ReCAP_sampler(ChicagoDeerdata$Harv.data,ChicagoDeerdata$Aeri.data
    ,ChicagoDeerdata$nage,ChicagoDeerdata$mean.f,ChicagoDeerdata$mean.s
    ,ChicagoDeerdata$mean.SRB,ChicagoDeerdata$mean.H,ChicagoDeerdata$mean.A
    ,n.iter = 50,burn.in = 50,thin.by = 1
    ,Assumptions = ChicagoDeerdata$Assumptions)


## another example
### This makes a full prior harvest matrix and the model will run a fully age
### and time specific reconstruction of harvest rate together with other vital rates.


mean.H.full <- ChicagoDeerdata$Assumptions$Harv$age %*%
    as.matrix( ChicagoDeerdata$mean.H) %*%
    ChicagoDeerdata$Assumptions$Harv$time
set.seed(42)
ReCAP_sampler(ChicagoDeerdata$Harv.data,ChicagoDeerdata$Aeri.data
    ,ChicagoDeerdata$nage,ChicagoDeerdata$mean.f,ChicagoDeerdata$mean.s
    ,ChicagoDeerdata$mean.SRB,mean.H.full,ChicagoDeerdata$mean.A
    ,n.iter = 50,burn.in = 50,thin.by = 1
    )

YunyiShen/ReDDLeslie documentation built on April 1, 2022, 6:58 a.m.