Plot svyplan objects

Visualize sampling fractions per stratum or power curves from svyplan results.

Usage

# S3 method for class 'svyplan_strata'
plot(x, ...)

# S3 method for class 'svyplan_power'
plot(x, npoints = 101L, ...)

# S3 method for class 'svyplan_n'
plot(x, npoints = 25L, newdata = NULL, ...)

Arguments

x

A svyplan object.

...

Additional graphical parameters passed to barplot() (for strata) or plot() (for power and the budget frontier). These override the defaults, so you can set main, col, ylab, xlab, ylim, etc.

npoints

Number of points in the grid: the power curve (default 101) or the budget frontier (default 25, since each point is a solve).

newdata

Optional one-column data frame of budget values for plot.svyplan_n(). The default sweeps from the cheapest design that meets the hard targets up to twice the fitted budget.

Value

x, invisibly.

Details

plot.svyplan_strata() draws a bar chart of per-stratum sampling fractions (f = n / N) using barplot(). This shows how intensively each stratum is sampled, under Neyman allocation, high-variance strata get higher fractions. A dashed horizontal line marks the overall sampling fraction (n / N). Defaults: col = "grey40", ylab = "Sampling fraction (f)", las = 2.

plot.svyplan_power() draws the power-vs-sample-size curve using plot(). The solved point is shown as a filled dot, with dashed reference lines at the computed power and sample size, and a dotted line at the significance level. Defaults: ylim = c(0, 1), type = "l", xlab = "Sample size (per group)", ylab = "Power".

plot.svyplan_n() draws the budget frontier for a fixed-budget joint allocation (n_alloc() with objective and budget): what precision each budget buys on the objective indicator, over the range where the hard targets remain fundable. The fitted design is a filled dot. The curve is the same one predict() returns as a table, so read exact numbers there. Its shape is the point: the objective falls as 1 / cost, so the marginal return on budget flattens, and the plot shows where. Other svyplan_n results have no frontier to draw and produce an error naming what is plottable.

See also

predict.svyplan for the sensitivity grids these curves are drawn from, and strata_bound(), power_prop(), n_alloc() for the results that are plottable.

Examples

# Sampling fraction per stratum
set.seed(1907)
sb <- strata_bound(rlnorm(2000, 6, 1), n_strata = 4, n = 200,
                    method = "cumrootf")
plot(sb)


# Custom color
plot(sb, col = "steelblue")


# Power curve with defaults
pw <- power_prop(p1 = 0.30, p2 = 0.40, power = 0.80)
plot(pw)


# Custom line width and color
plot(pw, lwd = 2, col = "darkred")


# Budget frontier: what each budget buys on the objective indicator
frame <- data.frame(
  stratum = c("A", "B", "C"),
  N = c(4000, 3000, 3000),
  unit_cost = c(1, 1.2, 1.5)
)
measures <- data.frame(
  stratum = rep(frame$stratum, 2),
  name = rep(c("vaccination", "income"), each = 3),
  p = c(0.5, 0.4, 0.6, rep(NA, 3)),
  mean = c(rep(NA, 3), 50, 55, 60),
  sd = c(rep(NA, 3), 10, 12, 15)
)
targets <- data.frame(name = "vaccination", cv = 0.05)
fit <- n_alloc(frame, measures = measures, targets = targets,
               objective = "income", budget = 4000)
plot(fit)