Skip to contents

Picture a sampling design

Source: R/plot.R
plot.drawn_design.Rd

Draws the design against a population so you can see what it selects. Two views, and each answers a question that a table answers slowly:

Usage

# S3 method for class 'drawn_design'
plot(
  x,
  y,
  type = c("selection", "probability", "map"),
  seed = NULL,
  ncol = NULL,
  max_dots = 4000,
  main = NULL,
  palette = NULL,
  coords = NULL,
  ...
)

Arguments

x

A design object.

y

The population data frame to draw against.

type

"selection", "probability" or "map".

seed

Optional seed for the "selection" draw, so the picture is reproducible.

ncol

Dots per row in the "selection" grid. Defaults to whatever fills the panel at its current aspect ratio.

max_dots

Frames larger than this are shown as an evenly spaced subset, noted under the title. Keeps individual dots visible.

main

Title. Defaults to a description of the design.

palette

Named list overriding any of surface, ink, secondary, muted, recessive, accent, fill, rule.

coords

For type = "map", the two columns to place rows by, c(x, y). Defaults to the design's own across (design_spread()) or coords (design_spatial()); required for any other design.

...

Passed to the underlying plot call.

Value

x, invisibly. Called for the plot.

Details

"selection"

Every row of the frame as a dot, in frame order, with the selected ones filled in. Designs look distinct: simple random sampling scatters, systematic makes a lattice, cluster sampling takes solid contiguous runs, and probability-proportional-to-size thickens wherever the weight is large. If your frame is sorted by something meaningful, an unintended pattern shows up immediately.

"probability"

Each row's inclusion probability across the frame, as a step. Flat means every row had the same chance; plateaus mean strata; a rise means size-proportional selection. Rows the design can never reach sit at zero and are marked, which is usually the thing worth finding out.

"map"

Every row placed by two columns — coordinates, or any two numeric variables — with the selected ones filled in. This is the view for design_spread() and design_spatial(), where what matters is how evenly the sample covers the space: compare a simple random sample's clumps and gaps against a spread one.

Base graphics, so there is no plotting dependency to install.

See also

draw(), inclusion_prob()

Examples

pop <- data.frame(
  id = 1:400,
  site = rep(c("a", "b", "c", "d"), times = c(200, 100, 60, 40)),
  cl = rep(paste0("c", 1:40), each = 10)
)

op <- par(mfrow = c(2, 1))

# Scattered, versus a visible lattice
plot(design_simple(n = 60), pop, seed = 1)
plot(design_systematic(interval = 7), pop, seed = 1)


par(op)

# Where the probabilities sit
plot(design_stratified("site", n = 60), pop, type = "probability")