Swarm of every re-analysis against the filed result

Two forms. Single census (style = "bubbles", the default for one census): the bubble swarm of earlier versions, one row per census on a percent-of-filed axis. Several rows (style = "dots", the default for a list): one panel per endpoint-metric row, the filed estimate and its 95% interval drawn as the axis of the panel with the filed value labelled, every counted alternative analysis as a dot placed by its estimate and coloured by the decision it varies (statistical: covariance, inference, covariate set, weighting, grid, estimator; data-handling: data set, population, intercurrent-event strategy, imputation), the counted deletions as a strip of small light dots beneath the axis, a vertical line at no difference, and optional labelled reference lines, one value for every panel or one per panel. Each panel labels its own filed estimate and interval ends to three significant digits. Diagnostic refits (kind = "diagnostic", such as the complete-case refit of the standard plan) are drawn as hollow dots in the colour of their decision group; they are not counted and do not enter the verdict. Counted rows that return no estimate (rank tests) are not drawn; their number per row is in the caption attribute of the returned plot, with the counts of alternatives, deletions and diagnostic refits.

Usage

plot_swarm(
  x,
  style = c("auto", "bubbles", "dots"),
  reference = NULL,
  reference_label = NULL,
  scale = c("percent", "estimate"),
  digits = 2,
  r0 = 0.004,
  favours = NULL,
  xlab = NULL,
  xlim = NULL
)

Arguments

x: A stress_census, a stress_result (from run_plan()), the stress_results list of run_plans(), or a list of censuses or results (one row each). List names, plan titles or census labels name the rows.
style: "bubbles" or "dots"; "auto" chooses bubbles for one census and dots for a list.
reference: Optional x position of a reference line (a clinically meaningful difference, say). In the dots form either one value, drawn in every panel, or one value per row: a vector named by row, or unnamed in row order. NA leaves a panel without a line. Rows on different scales (points and percent change, say) need their own values.
reference_label: Text printed at the reference line: one string, or one per row in the same form as reference.
scale, digits, r0, favours, xlab, xlim: Settings of the bubble form (see the earlier documentation): axis scale, rounding for "the same value", bubble radius, side labels, axis label and limits.

Value

A ggplot object. In the dots form the attribute caption is a character vector, one entry per row, with the counts of alternatives drawn, deletions drawn and rows without an estimate.

Examples

spec <- stress_spec(mmrm::fev_data, "FEV1", "USUBJID", "AVISIT", "ARMCD", "PBO",
                    covariates = ~ RACE + SEX + FEV1_BL)
cen <- stress_census(spec, stress_covariance(spec, c("us", "cs", "ar1h")),
                     stress_covariates(spec), stress_loo(spec, by = "RACE"))
if (requireNamespace("ggplot2", quietly = TRUE)) {
  plot_swarm(cen)                                           # one census: bubbles
  spec2 <- stress_spec(mmrm::fev_data, "FEV1", "USUBJID", "AVISIT", "ARMCD", "PBO",
                       covariates = ~ FEV1_BL, contrast_visit = "VIS3")
  cen2 <- stress_census(spec2, stress_covariance(spec2, c("us", "cs")),
                        stress_loo(spec2, by = "SEX"))
  g <- plot_swarm(list("FEV1 at visit 4" = cen, "FEV1 at visit 3" = cen2), reference = 1,
                  reference_label = "1 unit")
  attr(g, "caption")
  plot_swarm(list("FEV1 at visit 4" = cen, "FEV1 at visit 3" = cen2),     # one line per panel
             reference = c("FEV1 at visit 4" = 1, "FEV1 at visit 3" = 2),
             reference_label = c("1 unit", "2 units"))
}