Lie factor

The lie factor divides the proportional change shown in a graphic by the proportional change in the data. A value of one means those changes agree. Tufte treats values outside roughly 0.95 to 1.05 as distortion.

Usage

lie_factor(x, ...)

# S3 method for class 'numeric'
lie_factor(x, graphic, ...)

# S3 method for class 'ggplot'
lie_factor(x, ...)

# Default S3 method
lie_factor(x, ...)

Arguments

x

Either a numeric vector of underlying data values, or a ggplot object.

...

Unused.

graphic

For the numeric method, a numeric vector of the same length giving the sizes drawn in the graphic.

Value

A numeric lie factor, or NA when there's nothing to compare. The ggplot method returns 1 for a plot with no bars or with a zero baseline. For supported Cartesian bars it checks each panel and returns the largest distortion, including negative and reversed axes. Nonlinear coordinates and truncated stacked or floating bars return NA when no supported comparison is available.

Details

Supply two numeric vectors to compare data values with the sizes drawn. Or supply a ggplot with bars to measure the effect of a non-zero baseline. The plot method is limited to supported bar comparisons; a result of one isn't a general assessment of the figure's accuracy.

Examples

# Tufte's fuel-economy example: an 18 percent change drawn as 783 percent.
lie_factor(c(18.0, 27.5), c(0.6, 5.3))
#> [1] 14.84211

library(ggplot2)
d <- data.frame(g = c("a", "b"), v = c(100, 110))
lie_factor(ggplot(d, aes(g, v)) + geom_col() +
             coord_cartesian(ylim = c(95, 115)))
#> [1] 16.66667