Compare estimation error across designs (single replicate)

Plots estimation error of the chosen targets for two or more study designs, each represented by a single md_run() output. Use this function to quickly compare how design choices affect estimation performance before committing to a full replication with md_replicate().

Because each design is represented by a single stochastic run, results are preliminary. For robust, publication-ready comparisons, run md_replicate() for each design and compare with md_compare().

Usage

md_compare_preview(
  ...,
  n_resamples = NULL,
  error_threshold = 0.05,
  pal = c("#007d80", "#A12C3B")
)

Arguments

...

One or more movedesign_processed objects, each returned by md_run(), or a single list containing such objects. Each element represents one study design to compare. Designs typically differ in sampling parameters such as dur, dti, or n_individuals, but any valid inputs can be compared.

n_resamples

A single positive integer. The number of random combinations of individuals generated at each population sample size per design. Each combination produces one population-level estimate. Set to NULL to plot raw estimates without resampling.

error_threshold

Numeric. Relative error threshold shown as a horizontal reference line in the plot (e.g. 0.05 for 5%).

pal

A character vector of two colors, used for estimates within and outside the error threshold respectively. Defaults to c("#007d80", "#A12C3B").

Value

A ggplot object. Displays relative error as a function of population sample size, with one panel (or two if two target) per design. Point estimates, confidence intervals, and a horizontal reference line at error_threshold.

Details

If n_resamples is not NULL, the function draws n_resamples random combinations of individuals at each population sample size and computes a population-level estimate for each. This step

Each design is represented by a single stochastic run. Apparent differences between designs may reflect random variation rather than genuine performance differences. Use md_replicate() to generate robust, replicated results for each design, and md_compare() to compare multiple designs.

See also

md_run() to generate each input object. md_plot_preview() for a single-design equivalent. md_replicate() for robust multi-replicate outputs per design. md_check() to assess convergence across replicates.

Examples

if (interactive()) {

  data(buffalo)
  inputA <- md_prepare(
    data = buffalo,
    models = models,
    species = "buffalo",
    n_individuals = 5,
    dur = list(value = 1, unit = "month"),
    dti = list(value = 1, unit = "day"),
    add_individual_variation = FALSE,
    grouped = TRUE,
    set_target = "hr",
    which_meta = "mean")
  
  inputB <- md_prepare(
    data = buffalo,
    models = models,
    species = "buffalo",
    n_individuals = 5,
    dur = list(value = 10, unit = "days"),
    dti = list(value = 1, unit = "day"),
    add_individual_variation = TRUE,
    grouped = TRUE,
    set_target = "hr",
    which_meta = "mean")
  
  outputA <- md_run(inputA)
  outputB <- md_run(inputB)
  md_compare_preview(list(outputA,
                          outputB), error_threshold = 0.05)
}