Add a Sensitivity/Specificity Box

Add a shaded box, typically to a ROC curve, that corresponds to the 95% joint binomial confidence interval of the sensitivity and specificity.

Calculate the joint 95% confidence interval given sensitivity and specificity.

Usage

add_ss_box(x, col = "black", alpha = 0.35)

calc_joint_CI95(sens, spec, n.controls, n.cases)

Arguments

x: A 2x2 data frame or tibble containing the lower and upper CI95 joint confidence limits for sensitivity and specificity. A call to calc_joint_CI95() generates values in this specified format.
col: character(1) or integer(1). Specify the colors for lines, points, bar, box, or ROC.
alpha: numeric(1) in [0, 1]. The color transparency. See also ggplot2::alpha().
sens: numeric(n). The sensitivity: \([0, 1]\).
spec: numeric(n). The specificity: \([0, 1]\).
n.controls: integer(1). Number of control or non-cases.
n.cases: integer(1). Number of cases/disease.

Value

A \(2x2\) matrix containing rows of sensitivity and specificity respectively and columns of lower and upper 95% joint confidence intervals respectively.

Details

Recall that the ROC curve is 1 - specificity, therefore the added box involves internally inverting the specificity limits so that the interval matches the plot.

Author

Stu Field, Amanda Hiser

Mike Mehan

Examples

g <- ggplot2::ggplot(data.frame(x = 0.2, y = 0.8), ggplot2::aes(x = x, y = y)) +
  ggplot2::geom_point(shape = 18, size = 3) +
  ggplot2::lims(x = 0:1, y = 0:1) +
  ggplot2::labs(y = "Sensitivity", x = "1 - Specificity")
g


# calculate CI95s for 80/80 sens/spec
ci95 <- calc_joint_CI95(0.8, 0.8, 35, 65)
ci95
#> # A tibble: 2 × 2
#>   lower upper
#> * <dbl> <dbl>
#> 1 0.689 0.911
#> 2 0.649 0.951

# unequal box due to class imbalance (65/35)
g + add_ss_box(ci95, col = "blue", alpha = 0.25)