Movatterモバイル変換

Title:	Presence-Only for Marked Point Process
Version:	0.1.3
Date:	2022-12-12
Description:	Inspired by Moreira and Gamerman (2022) <doi:10.1214/21-AOAS1569>, this methodology expands the idea by including Marks in the point process. Using efficient 'C++' code, the estimation is possible and made faster with 'OpenMP'https://www.openmp.org/ enabled computers. This package was developed under the project PTDC/MAT-STA/28243/2017, supported by Portuguese funds through the Portuguese Foundation for Science and Technology (FCT).
License:	GPL (≥ 3)
Encoding:	UTF-8
RoxygenNote:	7.2.0
Depends:	R (≥ 2.14.0)
LinkingTo:	Rcpp, RcppEigen, RcppProgress
Imports:	Rcpp, coda, geoR, parallel, methods, graphics, stats, tools
Suggests:	bayesplot, ggplot2, MASS
Collate:	'RcppExports.R' 'covariance_importance.R' 'prior-class.R''initial-class.R' 'model-class.R' 'fit-class.R''pompp-package.R' 'pompp.R'
NeedsCompilation:	yes
Packaged:	2022-12-12 17:11:54 UTC; anthorg
Author:	Guido Alberti Moreira [cre, aut]
Maintainer:	Guido Alberti Moreira <guidoalber@gmail.com>
Repository:	CRAN
Date/Publication:	2022-12-12 23:50:05 UTC

Generic class for the beta and delta parameters.

Description

Generic class for the beta and delta parameters.

Usage

## S4 method for signature 'BetaDeltaPrior'show(object)## S4 method for signature 'BetaDeltaPrior'print(x, ...)## S3 method for class 'BetaDeltaPrior'print(x, ...)

Arguments

Value

show andprint: The invisible object.

Fields

family

The family of distributions of the prior.

Create a Gamma prior object for model specification.

Description

Constructor forGammaPrior-class objects

Usage

GammaPrior(shape, rate)

Arguments

Value

AGammaPrior object with adequate slots.

Gamma prior class for the LambdaStar parameter.

Description

This is used to represent the prior for lambdaStar individually. Itstill needs to be joined with the prior for Beta and Delta to be usedin a model.

Usage

## S4 method for signature 'GammaPrior'names(x)## S4 method for signature 'GammaPrior'x$name## S4 replacement method for signature 'GammaPrior'x$name <- value## S4 method for signature 'GammaPrior'show(object)## S4 method for signature 'GammaPrior'print(x, ...)## S3 method for class 'GammaPrior'print(x, ...)

Arguments

Value

names: A character vector with the prior parameters.

`$` The requested slot's value.

`$<-`: The new object with the updated slot.

show andprint: The invisible object.

Fields

shape

The shape parameter of the Gamma distribution.

rate

The rate parameter of the Gamma distribution.

See Also

prior

Examples

GammaPrior(0.0001, 0.0001)

Generic class for the LambdaStar parameters.

Description

Generic class for the LambdaStar parameters.

Usage

## S4 method for signature 'LambdaStarPrior'show(object)

Arguments

Value

show andprint: The invisible object.

Fields

family

The family of distributions of the prior.

Create a Normal prior object for model specification.

Description

Constructor forNormalPrior-class objects

Usage

NormalPrior(mu, Sigma)

Arguments

Details

Matrix Sigma must be square and positive definite. Its dimensionsmust match mu's length.

Value

ANormalPrior object with adequate slots.

See Also

prior

Examples

NormalPrior(rep(0, 10), diag(10) * 10)

Normal prior class for Beta and Delta parameters.

Description

This is used to represent the prior for Beta and Delta individually. Theystill need to be joined to be used in a model.

Usage

## S4 method for signature 'NormalPrior'names(x)## S4 method for signature 'NormalPrior'x$name## S4 replacement method for signature 'NormalPrior'x$name <- value## S4 method for signature 'NormalPrior'show(object)## S4 method for signature 'NormalPrior'print(x, ...)## S3 method for class 'NormalPrior'print(x, ...)

Arguments

Value

names: A character vector with the prior parameters.

`$`: The requested slot's value.

`$<-`: The new object with the updated slot.

show andprint: The invisible object.

Fields

mu

The mean vector for the prior.

Sigma

The covariance matrix for the prior.

Class for covariates importance matrices

Description

Objects of this class is the output of the "covariates_importance" objectfrom thepompp_fit-class. It can be plotted which usesthegraphics package. Theprint methodgives a point-wise estimation, the same seen in thebacplot method.Bothplot andboxplot methods use the posterior distributionof the importance.

Usage

## S3 method for class 'covariates_importance'print(x, component = "intensity", ...)## S3 method for class 'covariates_importance'plot(  x,  component = "intensity",  y = "importance",  quantiles = c(0.025, 0.5, 0.975),  ...)## S3 method for class 'covariates_importance'barplot(height, component = "intensity", y, ...)## S3 method for class 'covariates_importance'boxplot(x, component = "intensity", ...)

Arguments

Details

Objects of this class have two matrices where the Monte Carlo samples on therows and parameters on the columns. One matrix is for the intensityimportance and the other for the observability importance.

Value

The invisible object.

Nothing is returned. Plot is called and drawn on the configureddevice.

A barplot. Seebarplot for details. If component is selectedas"both", only the second barplot is returned.

A boxplot. Seeboxplot for details. If component is selectedas"both", only the second boxplot is returned.

See Also

barplot.

boxplot.

Fit presence-only data using the Presence-Only for Marked Point Process model

Description

The model uses a data augmentation scheme to avoid performing approximationson the likelihood function.

Usage

fit_pompp(  object,  background,  mcmc_setup = list(iter = 5000),  verbose = TRUE,  ...)## S4 method for signature 'pompp_model,matrix'fit_pompp(  object,  background,  neighborhoodSize = 20,  mcmc_setup,  verbose = TRUE,  area = 1,  cores = parallel::detectCores(),  ...)checkFormatBackground(object, background)

Arguments

Details

The background is kept outside of the

Value

An object of class"pompp_fit".

See Also

pompp_model andpompp_fit-class.

Examples

beta <- c(-1, 2) # Intercept = -1. Only one covariatedelta <- c(3, 4) # Intercept = 3. Only one covariatelambdaStar <- 1000gamma <- 2mu <- 5total_points <- rpois(1, lambdaStar)random_points <- cbind(runif(total_points), runif(total_points))grid_size <- 50# Find covariate values to explain the species occurrence.# We give them a Gaussian spatial structure.reg_grid <- as.matrix(expand.grid(seq(0, 1, len = grid_size), seq(0, 1, len = grid_size)))Z <- MASS::mvrnorm(1, rep(0, total_points + grid_size * grid_size),  3 * exp(-as.matrix(dist(rbind(random_points, reg_grid))) / 0.2))Z1 <- Z[1:total_points]; Z2 <- Z[-(1:total_points)]# Thin the points by comparing the retaining probabilities with uniforms# in the log scale to find the occurrencesoccurrences <- log(runif(total_points)) <= -log1p(exp(-beta[1] - beta[2] * Z1))n_occurrences <- sum(occurrences)occurrences_points <- random_points[occurrences,]occurrences_Z <- Z1[occurrences]# Find covariate values to explain the observation bias.# Additionally create a regular grid to plot the covariate later.W <- MASS::mvrnorm(1, rep(0, n_occurrences + grid_size * grid_size),  2 * exp(-as.matrix(dist(rbind(occurrences_points, reg_grid))) / 0.3))W1 <- W[1:n_occurrences]; W2 <- W[-(1:n_occurrences)]S <- MASS::mvrnorm(1, rep(0, n_occurrences),  2 * exp(-as.matrix(dist(occurrences_points)) / 0.3))# Find the presence-only observations.marks <- exp(mu + S + rnorm(n_occurrences, 0, 0.3))po_sightings <- log(runif(n_occurrences)) <= -log1p(exp(-delta[1] - delta[2] * W1 - gamma * S))n_po <- sum(po_sightings)po_points <- occurrences_points[po_sightings, ]po_Z <- occurrences_Z[po_sightings]po_W <- W1[po_sightings]po_marks <- marks[po_sightings]jointPrior <- prior(  NormalPrior(rep(0, 2), 10 * diag(2)), # BetaNormalPrior(rep(0, 3), 10 * diag(3)), # DeltaGammaPrior(0.00001, 0.00001), # LambdaStarNormalPrior(0, 100), GammaPrior(0.001, 0.001) # Marks)model <- pompp_model(po = cbind(po_Z, po_W, po_points, po_marks),intensitySelection = 1, observabilitySelection = 2, marksSelection = 5,                    coordinates = 3:4,                    intensityLink = "logit", observabilityLink = "logit",                    initial_values = 2, joint_prior = jointPrior)bkg <- cbind(Z2, W2, reg_grid) # Create background# Be prepared to wait a long time for thisfit <- fit_pompp(model, bkg, neighborhoodSize = 20, area = 1,  mcmc_setup = list(burnin = 1000, iter = 2000), cores = 1)summary(fit)# Rhat upper CI values are above 1.1. More iterations are needed...

Initial values constructor for pompp modeling

Description

Helper function to create a valid set of initial values to be used with thefit_pompp function.

Usage

initial(  beta = numeric(),  delta = numeric(),  lambdaStar = numeric(),  marksMean = numeric(),  marksPrecision = numeric(),  random = FALSE)

Arguments

Value

Apompp_initial object. It can be used in thefit_pompp function by itself, but must be in a list if multipleinitial values are supplied. Initial values can be combined by adding them(with the use of+).

See Also

pompp_initial-class.

Examples

# Let us create initial values for a model with, say, 3 intensity covariates# and 4 observability covariates. We add an initial values for both these# cases due to the intercepts.# This first one isin1 <- initial(rep(0, 4), c(0, 2, -1, -2, 3), 100, 0, 1)# Then we initalize some randomly.in2 <- initial(4, 5, 100, 0, 1, random = TRUE)# We can even multiply the random one to generate more. Let us join them all# to include in a model.initial_values <- in1 + in2 * 3# 4 chains are initialized.

Class for the result of the MCMC procedure.

Description

Objects of this class are the main objects of this package. They containmuch information about the fitted model.

Usage

## S4 method for signature 'pompp_fit'show(object)## S4 method for signature 'pompp_fit'print(x, ...)## S3 method for class 'pompp_fit'print(x, ...)## S4 method for signature 'pompp_fit'summary(object, ...)## S3 method for class 'pompp_fit'summary(object, ...)## S4 method for signature 'pompp_fit'names(x)## S3 method for class 'pompp_fit'names(x)## S4 method for signature 'pompp_fit,ANY,ANY'x[[i]]## S4 method for signature 'pompp_fit'x$name## S4 method for signature 'pompp_fit'as.array(x, ...)## S3 method for class 'pompp_fit'as.array(x, ...)## S4 method for signature 'pompp_fit'as.matrix(x, ...)## S3 method for class 'pompp_fit'as.matrix(x, ...)## S4 method for signature 'pompp_fit'as.data.frame(x, row.names = NULL, optional = FALSE, ...)## S3 method for class 'pompp_fit'as.data.frame(x, row.names = NULL, optional = FALSE, ...)## S4 method for signature 'pompp_fit,pompp_fit'e1 + e2## S4 method for signature 'pompp_fit'c(x, ...)

Arguments

Value

show andprint: The invisible object.

summary: A matrix with the summary statistics of thefit. It is also printed in theprint method. The summary can betreated as a matrix, such as retrieving rows/columns and creating tableswith thextable package.

names: A character vector with the available optionsfor the`$` and`[[` methods.

`$` and`[[`: The requested slot.Available options are not necessarily the class slots, and can be checkedwith thenames method.

as.array: Anarray with dimensions I x C x P,where I stands for number of iterations, C for number of chains and P fortotal number of parameters. P is actually larger than the number ofparameters in the model, as the the generated sizes of the latent processesand the log-posterior are also included. This is organized so that is readyfor thebayesplot package functions.

as.matrix: The dimension of the output isI * C x (P + 2), where I stands for number of iterations, C for number ofchains and P for total number of parameters. P is actually larger than thenumber of parameters in the model, as the generated sizes of the latentprocesses and the log-posterior are also included.

Two extra columns are included to indicate to which chain and to whichiteration that draw belongs.

as.data.frame: The dimension of the output isI*C x P + 2, where I stands for number of iterations, C for number of chainsand P for total number of parameters. P is actually larger than the numberof parameters in the model, as the generated sizes of the latent processesand the log-posterior are also included.

Two extra columns are included to indicate to which chain and to whichiteration that draw belongs. This is to facilitate the use of plottingresults via theggplot2 package if desired.

Ifrow.names is left atNULL then row names are created asCcIi where c is the chain and i is the iteration of that row.

+: A newpompp_fit object where the chainsare combined into a new multi-chain object. This can be used if chains arerun in separate occasions or computers to combine them into a single objectfor analysis.

c: A newpompp_fit object where the chainsare combined into a new multi-chain object. The+ method isused for that, with the same arguments restrictions and results.

Fields

fit

The actual fit from the model. It is an object of classmcmc.list, as generated from thecoda package.

original

The model used to generate the chains, an object with classpompp_model.

backgroundSummary

A small summary of the original backgroundcovariates. This is to ensure that continuing the chains will use theidentical background matrix. Only the summary is kept for storage efficiency.

area

A positive number indicating the area measure of the region beingstudied.

parnames

The names of the parameters. If the model used selects thecovariates with column names, they are replicated here. If they are thecolumn indexes, names are generated for identification.

mcmc_setup

The original mcmc setup used.

See Also

fit_pompp

Class for the initial values for the MCMC for the pompp package

Description

Class for the initial values for the MCMC for the pompp package

Usage

## S4 method for signature 'pompp_initial'names(x)## S4 method for signature 'pompp_initial'x$name## S4 method for signature 'pompp_initial,ANY'e1 + e2## S4 method for signature 'list,pompp_initial'e1 + e2## S4 method for signature 'pompp_initial,list'e1 + e2## S4 method for signature 'pompp_initial,numeric'e1 * e2## S4 method for signature 'numeric,pompp_initial'e1 * e2## S4 method for signature 'pompp_initial'show(object)## S4 method for signature 'pompp_initial'print(x, ...)## S3 method for class 'pompp_initial'print(x, ...)

Arguments

Value

names: A character vector with the initializedparameter names.

`$`: The requested initial value (in case ofLambdaStar) or values (in case of Beta or Delta).

+: A list with the objects. Useful to start thefit_pompp function, as it requires a list of initial values.

*: A list withe2 random initial values.

show andprint: The invisible object.

Fields

beta

Initial values for beta.

delta

Initial values for delta.

lambdaStar

Initial values for lambdaStar.

tag

Indicates the source of the initial values.

Build a model to be used in thepompp fitting function

Description

Constructor forpompp_model-class objects, built to facilitatethe use of the fitting function. The output of this function has thenecessary signature for the fit_pompp function to start the model fit.

Usage

pompp_model(  po,  intensitySelection,  observabilitySelection,  marksSelection,  coordinates,  intensityLink = "logit",  observabilityLink = "logit",  initial_values = 1,  joint_prior = prior(beta = NormalPrior(rep(0, length(intensitySelection) + 1), 10 *    diag(length(intensitySelection) + 1)), delta = NormalPrior(rep(0,    length(observabilitySelection) + 1), 10 * diag(length(observabilitySelection) + 1)),    lambdaStar = GammaPrior(1e-10, 1e-10), marksMean = NormalPrior(1, 100),    marksPrecision = GammaPrior(0.001, 0.001)),  verbose = TRUE)

Arguments

Value

Apompp_model object with the requested slots. It is readyto be used in thefit_pompp function.

See Also

initial,prior andfit_pompp.

Examples

# Let us simulate some data to showcase the creation of the model.beta <- c(-1, 2)delta <- c(3, 4)lambdaStar <- 1000gamma <- 2mu <- 5total_points <- rpois(1, lambdaStar)random_points <- cbind(runif(total_points), runif(total_points))# Find covariate values to explain the species occurrence.# We give them a Gaussian spatial structure.Z <- MASS::mvrnorm(1, rep(0, total_points), 3 * exp(-as.matrix(dist(random_points)) / 0.2))# Thin the points by comparing the retaining probabilities with uniforms# in the log scale to find the occurrencesoccurrences <- log(runif(total_points)) <= -log1p(exp(-beta[1] - beta[2] * Z))n_occurrences <- sum(occurrences)occurrences_points <- random_points[occurrences,]occurrences_Z <- Z[occurrences]# Find covariate values to explain the observation bias.# Additionally create a regular grid to plot the covariate later.W <- MASS::mvrnorm(1, rep(0, n_occurrences), 2 * exp(-as.matrix(dist(occurrences_points)) / 0.3))S <- MASS::mvrnorm(1, rep(0, n_occurrences), 2 * exp(-as.matrix(dist(occurrences_points)) / 0.3))# Find the presence-only observations.marks <- exp(mu + S + rnorm(n_occurrences, 0, 0.3))po_sightings <- log(runif(n_occurrences)) <= -log1p(exp(-delta[1] - delta[2] * W - gamma * S))n_po <- sum(po_sightings)po_points <- occurrences_points[po_sightings, ]po_Z <- occurrences_Z[po_sightings]po_W <- W[po_sightings]po_marks <- marks[po_sightings]# Now we create the modelmodel <- pompp_model(po = cbind(po_Z, po_W, po_points, po_marks),  intensitySelection = 1, observabilitySelection = 2,  marksSelection = 5, coordinates = 3:4,  intensityLink = "logit", observabilityLink = "logit",  initial_values = 2, joint_prior = prior(    NormalPrior(rep(0, 2), 10 * diag(2)),    NormalPrior(rep(0, 3), 10 * diag(3)),    GammaPrior(1e-4, 1e-4),    NormalPrior(0, 100), GammaPrior(0.001, 0.001)))# Check how it is.model

Class that defines a model for the pompp package.

Description

The model includes the presence-only data, all selected variables, the linkfunctions forq andp, the initial values and the priordistribution.

Usage

## S4 method for signature 'pompp_model'names(x)## S4 method for signature 'pompp_model'x$name## S4 replacement method for signature 'pompp_model'x$name <- value## S4 method for signature 'pompp_model'show(object)## S4 method for signature 'pompp_model'print(x, ...)## S3 method for class 'pompp_model'print(x, ...)

Arguments

Value

names: A character vector with possible optionsfor the`$` and`$<-` methods.

`$`: The requested slot's value.

`$<-`: The new object with the updated slot.

show andprint: The invisible object.

Fields

po

The matrix containing the covariates values for the data.

intensityLink

A string informing about the chosen link for theintensity covariates. Current acceptable choice is only"logit".

intensitySelection

A vector containing the indexes of the selectedintensity columns in thepo matrix.

observabilityLink

A string informing about the chosen link for theobservability covariates. Current acceptable choice is only"logit".

observabilitySelection

A vector containing the indexes of the selectedobservability columns in thepo matrix.

marksSelection

A single value containing the index of the selectedmarks column in thepo matrix.

coordinates

A vector of two values containing the column positionsof the longitude and latitude in thepo matrix.

init

A list with objects of classpompp_initial indicatingthe initial values for each chain. The length of this list tells the programhow many chains are requested to be run.

prior

An object of classpompp_prior which indicates thejoint prior distribution for the model parameters.

iSelectedColumns

If the intensity covariates selection was made withthe name of the columns, they are stored in this slot.

oSelectedColumns

If the observability covariates selection was madewith the name of the columns, they are stored in this slot.

mSelectedColumns

If the marks selection was madewith the name of the column, it is stored in this slot.

See Also

pompp_initial-class andpompp_prior-class andpompp_model

Joint prior class for the pompp package parameters

Description

Objects of this class are the joining of independent priors for Beta, Deltaand LambdaStar. They can be used in thefit_pompp function.

Usage

## S4 method for signature 'pompp_prior'names(x)## S4 method for signature 'pompp_prior'x$name## S4 method for signature 'pompp_prior'show(object)## S4 method for signature 'pompp_prior'print(x, ...)## S3 method for class 'pompp_prior'print(x, ...)## S4 method for signature 'pompp_prior'x$name## S4 replacement method for signature 'pompp_prior'x$name <- value

Arguments

Value

names: A character vector with the model parametersnames.

`$`: The requested slot's value.

`$<-`: The new object with the updated slot.

Fields

beta

An object of a class which inherits theBetaDeltaPrior S4class with the appropriate Beta prior.

delta

An object of a class which inherits theBetaDeltaPrior S4class with the appropriate Delta prior.

lambdaStar

An object of a class which inherits theLambdaStarPrior S4 class with the appropriate LambdaStar prior.

marksMean

An object of S4 classNormalPrior with the chosenprior for the marks mean

marksPrecision

An object of S4 classGammaPrior with the chosenprior for the marks precision

Build a joint prior for pompp model parameters

Description

Constructor forpompp_prior objects, which is used in thepompp_fit function. The generated prior is so that Beta, Deltaand LambdaStar are indepdendent a priori.

Usage

prior(beta, delta, lambdaStar, marksMean, marksPrecision)

Arguments

Value

Apompp_prior object with the adequate slots. It is ready tobe included in a model via thepompp_model function.

See Also

fit_pompp,NormalPrior,GammaPrior andpompp_model.

Examples

# Let us say there are 3 intensity covariates and 4 observability covariates.# One more element is included in both sets due to the intercepts.new_prior <- prior(  NormalPrior(rep(0, 4), 10 * diag(4)),  NormalPrior(rep(0, 5), 10 * diag(5)),  GammaPrior(0.0001, 0.0001),  NormalPrior(0, 100), GammaPrior(0.001, 0.001))