Half-width of a group's error bar

What linePlot() draws either side of a group's mean for error.type = "sd", "sem" or "ci95". Missing values are dropped before n is counted, so the bar describes the values that went into the mean.

Usage

error_bar_halfwidth(
  x,
  type = c("sd", "sem", "ci95"),
  ci.method = c("normal", "t")
)

Arguments

x

Numeric vector of one group's y-values.

type

One of "sd" (standard deviation), "sem" (standard error of the mean, sd / sqrt(n)) or "ci95" (95% confidence interval for the mean).

ci.method

For type = "ci95", "normal" scales the standard error by the 97.5th percentile of the normal distribution (1.96) and "t" by that of the t distribution on n - 1 degrees of freedom. Ignored for the other types.

Value

A single number, or NA_real_ for fewer than two non-missing values, which have no spread to draw.

Details

A module that summarises its own groups (over a numeric x, say, which linePlot's own summaries do not cover) can work out each interval with this and hand linePlot() the bounds through error.type = "columns", so its bars and ribbons match the line plot module's.

See also

linePlot()

Author

Jared Andrews

Examples

x <- c(4.1, 5.3, 4.8, 6.0, 5.5)
error_bar_halfwidth(x, "sd")
#> [1] 0.7231874
error_bar_halfwidth(x, "sem")
#> [1] 0.3234192
# The t interval is wider for small groups.
error_bar_halfwidth(x, "ci95", "normal")
#> [1] 0.63389
error_bar_halfwidth(x, "ci95", "t")
#> [1] 0.8979557