Package 'BayesianNetworks'

Check fitted model

Description

Check fitted model

Usage

check_model(fit = NULL, data = NULL)

Arguments

fit	A fitted model, as obtained from fit_model().
data	Data list (from prepare_data()).

Value

Model checks on console and graphical window.

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
check_model(fit, data = data)

---

Fit model

Description

Fit model

Usage

fit_model(
  data = NULL,
  model = c("sampling_effort", "Young2021", "varying_preferences"),
  beta = 0.01,
  ...
)

Arguments

data	A named list containing the required data, as obtained from prepare_data().
model	character. One of "Young2021", "sampling_effort", or "varying_preferences", or a path to a file describing the Stan model in case you want to use a modified Stan model.
beta	Rate of exponential prior on r (preference) parameter. Default beta is 0.01. Increase it if you have large count numbers (can examine the resultant prior using plot_prior()).
...	Further arguments for cmdstanr::sample(), like iter_warmup, iter_sampling, or thin, among others. It is recommended to increase the number of iterations (e.g. iter_sampling = 10000).

Value

A fitted model (cmdstanr::CmdStanMCMC() object).

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt)

---

Get posterior values

Description

Get posterior values

Usage

get_posterior(
  fit = NULL,
  data = NULL,
  param = c("all", "connectance", "preference", "plant.abund", "animal.abund",
    "int.prob", "link")
)

Arguments

fit	Fitted model (from fit_model())
data	Data list (from prepare_data())
param	character. Name of the parameter to retrieve the posterior samples.

Value

A data frame

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
get_posterior(fit, dt, param = "connectance")

int.prob <- get_posterior(fit, dt, param = "int.prob")
int.prob
int.prob |> tidybayes::mean_qi()  # mean edge probability

# all posteriors
get_posterior(fit, dt, param = "all")

---

Get seed used to fit a model

Description

Get seed used to fit a model

Usage

get_seed(fit = NULL)

Arguments

fit	A fitted model

Value

A number

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
get_seed(fit)

---

Plot heatmap of observed counts

Description

Plot heatmap of observed counts

Usage

plot_counts_obs(mat = NULL, ...)

Arguments

mat	A matrix with count data reporting interaction frequency (e.g. visits to flowers, number of fruits consumed per plant or species). Plants must be in rows, Animals must be in columns.
...	Further arguments for network.tools::plot_web_heatmap().

Value

A ggplot object

Examples

data(web)
plot_counts_obs(web)
plot_counts_obs(web, sort = FALSE)
plot_counts_obs(web, zero.na = FALSE, sort = FALSE)

---

Plot heatmap of predicted counts

Description

Plot heatmap of predicted counts

Usage

plot_counts_pred(pred.df = NULL, ...)

Arguments

pred.df	A data frame containing the predicted counts, as generated by predict_counts()
...	Further arguments for network.tools::plot_web_heatmap().

Value

A ggplot object

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
pred.df <- predict_counts(fit, dt)
plot_counts_pred(pred.df)
plot_counts_pred(pred.df, sort = FALSE)

---

Plot predicted versus observed counts

Description

Plot predicted versus observed counts

Usage

plot_counts_pred_obs(
  pred.df = NULL,
  data = NULL,
  byplant = FALSE,
  width = 0.95,
  ...
)

Arguments

pred.df	A data frame containing the predicted counts, as generated by predict_counts()
data	Data list (from prepare_data())
byplant	Logical. If TRUE, show predicted and observed counts per plant (using ggplot2::facet_wrap()). If FALSE, show all interactions in the same plot.
width	width of the credible interval (default = 0.95).
...	Further arguments to be passed to ggplot2::facet_wrap() if byplant = TRUE, or to tidybayes::geom_pointinterval() if byplant = FALSE.

Value

A ggplot object

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
pred.df <- predict_counts(fit, dt)
plot_counts_pred_obs(pred.df, dt)
plot_counts_pred_obs(pred.df, dt, fatten_point = 3)
plot_counts_pred_obs(pred.df, dt, byplant = TRUE)
plot_counts_pred_obs(pred.df, dt, byplant = TRUE, scale = "free")

---

Plot heatmap of interaction probabilities

Description

Plot a heatmap of average interaction probabilities

Usage

plot_interaction_prob(post = NULL)

Arguments

post	Data frame containing the posterior probabilities, as generated from get_posterior().

Value

A ggplot object

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
post <- get_posterior(fit, dt)
plot_interaction_prob(post)

---

Plot prior distribution for r (preference) parameter

Description

The r (preference) parameter in Young et al. model takes a prior exponential distribution with rate = beta. Use this function to visualise the prior distribution of r given the chosen beta. Alternatively, if providing the fitted model, a plot comparing the prior versus posterior preference(s) will be produced.

Usage

plot_prior(beta = NULL, fit = NULL, data = NULL)

Arguments

beta	A number > 0. Rate of the exponential distribution.
fit	A fitted model, as obtained from fit_model().
data	Data list (from prepare_data()).

Value

A plot

Examples

## Providing value for beta
plot_prior(beta = 0.01)
plot_prior(beta = 0.001)

## Providing fitted model
data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
plot_prior(fit = fit, data = dt)

---

Plot heatmap of residuals

Description

Plot heatmap of residuals (observed - predicted counts).

Usage

plot_residuals(pred.df = NULL, data = NULL, ...)

Arguments

pred.df	A data frame containing the predicted counts, as generated by predict_counts()
data	Data list (from prepare_data())
...	Further arguments for network.tools::plot_web_heatmap().

Value

A ggplot object

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
pred.df <- predict_counts(fit, dt)
plot_residuals(pred.df, dt)
plot_residuals(pred.df, dt, sort = FALSE)

---

Predict interaction counts

Description

Generate the posterior predictive distribution of counts for every pairwise interaction.

Usage

predict_counts(fit = NULL, data = NULL)

Arguments

fit	Fitted model
data	Data list

Value

A data frame

Examples

data(web)
dt <- prepare_data(mat = web, sampl.eff = rep(20, nrow(web)))
fit <- fit_model(dt, refresh = 0)
predict_counts(fit, dt)

---

Prepare the data for modelling

Description

Prepare the data for modelling

Usage

prepare_data(mat = NULL, sampl.eff = NULL)

Arguments

mat	An integer matrix containing quantitative (not qualitative or binary) count data con interaction frequency (e.g. visits to flowers, number of fruits consumed per plant or species). Plants must be in rows, Animals must be in columns.
sampl.eff	A numeric vector with the sampling effort (e.g. observation hours) spent on each plant.

Value

A named list with all the data required to run the model.

Examples

data(web)
prepare_data(web, sampl.eff = rep(20, nrow(web)))

---

Plant-pollinator network

Description

An example bipartite network of 8 plant species and 21 pollinators, from Kaiser-Bunbury et al. 2017.

Usage

web

Format

web

A numeric (integer) matrix with 8 rows (representing plants) and 21 columns (representing animals)

Source

Kaiser-Bunbury, C., Mougal, J., Whittington, A. et al. Ecosystem restoration strengthens pollination network resilience and function. Nature 542, 223–227 (2017). https://doi.org/10.1038/nature21071

---