cosimmr_ffvb {cosimmr}

Run a cosimmr_input object through the Fixed Form Variational Bayes(FFVB) function

Description

This is the main function of cosimmr. It takes a cosimmr_input object created via cosimmr_load, runs it in fixed form Variational Bayes to determine the dietary proportions, and then outputs a cosimmr_output object for further analysis and plotting via plot.cosimmr_output.

Usage

cosimmr_ffvb(
  cosimmr_in,
  prior_control = list(mu_0 = rep(0, (cosimmr_in$n_sources * cosimmr_in$n_covariates)),
    mu_log_sig_sq_0 = rep(0, cosimmr_in$n_tracers), sigma_0 = 1, tau_shape = rep(1,
    cosimmr_in$n_tracers), tau_rate = rep(1, cosimmr_in$n_tracers)),
  ffvb_control = list(n_output = 3600, S = 500, P = 50, beta_1 = 0.75, beta_2 = 0.75, tau
    = 500, eps_0 = 0.0011, t_W = 500)
)

Arguments

cosimmr_in

An object created via the function cosimmr_load

prior_control

A list of values including arguments named mu_0 (prior for mu), and sigma_0 (prior for sigma).

ffvb_control

A list of values including arguments named n_output (number of rows in theta output), S (number of samples taken at each iteration of the algorithm), P (patience parameter), beta_1 and beta_2 (adaptive learning weights), tau (threshold for exploring learning space), eps_0 (fixed learning rate), t_W (rolling window size)

Value

An object of class cosimmr_output with two named top-level components:

input

The cosimmr_input object given to the cosimmr_ffvb function

output

A set of outputs produced by the FFVB function. These can be analysed using the summary.cosimmr_output and plot.cosimmr_output functions.

Author(s)

Emma Govan <[e-mail]>, Andrew Parnell

References

Andrew C. Parnell, Donald L. Phillips, Stuart Bearhop, Brice X. Semmens, Eric J. Ward, Jonathan W. Moore, Andrew L. Jackson, Jonathan Grey, David J. Kelly, and Richard Inger. Bayesian stable isotope mixing models. Environmetrics, 24(6):387–399, 2013.

Andrew C Parnell, Richard Inger, Stuart Bearhop, and Andrew L Jackson. Source partitioning using stable isotopes: coping with too much variation. PLoS ONE, 5(3):5, 2010.

See Also

cosimmr_load for creating objects suitable for this function, plot.cosimmr_input for creating isospace plots, summary.cosimmr_output for summarising output, and plot.cosimmr_output for plotting output.

Examples

## See the package vignette for a detailed run through of these examples

# Data set 1: 10 obs on 2 isos, 4 sources, with tefs and concdep
data(geese_data_day1)
x = c(1,2,3,2,1,3,2,1,2)
cosimmr_1 <- with(
  geese_data_day1,
  cosimmr_load(
    formula = mixtures ~ x,
    source_names = source_names,
    source_means = source_means,
    source_sds = source_sds,
    correction_means = correction_means,
    correction_sds = correction_sds,
    concentration_means = concentration_means
  )
)

# Plot
plot(cosimmr_1)

# Print
cosimmr_1

# FFVB run
cosimmr_1_out <- cosimmr_ffvb(cosimmr_1)

# Print it
print(cosimmr_1_out)

# Summary
summary(cosimmr_1_out, type = "correlations")
summary(cosimmr_1_out, type = "statistics")
ans <- summary(cosimmr_1_out, type = c("quantiles", "statistics"))

# Plot
plot(cosimmr_1_out, type = "beta_boxplot", cov_name = "x")
plot(cosimmr_1_out, type = "beta_histogram", cov_name = "x")