
Summarize and plot threshold-specific diagnostic accuracy
summary.mars_dta_thresholds.RdSummarize and plot threshold-specific diagnostic accuracy
Usage
# S3 method for class 'mars_dta_thresholds'
summary(object, level = 0.95, ...)
# S3 method for class 'mars_dta_thresholds'
predict(object, thresholds = NULL, level = 0.95, ...)
# S3 method for class 'mars_dta_thresholds'
plot(
x,
...,
xlim = c(0, 1),
ylim = c(0, 1),
xlab = "False-positive rate",
ylab = "Sensitivity",
labels = TRUE,
thresholds = NULL,
type = if (identical(x$family, "multinomial")) "l" else "p"
)Arguments
- object, x
A fitted
mars_dta_thresholds()object.- level
Confidence level between zero and one.
- ...
Graphical arguments passed to
graphics::plot()when plotting.- thresholds
Numeric vector of cutoffs for prediction or plotting. Defaults to the observed cutoffs for predictions and Gaussian plots. Multinomial plots default to a grid spanning the observed cutoffs. Gaussian predictions require reported cutoffs.
- xlim, ylim, xlab, ylab
Axis limits and labels for the ROC plot.
- labels
Logical. Label each operating point with its reported cutoff.
- type
Plot type passed to
graphics::plot(). Defaults to points for Gaussian fits and a line for multinomial fits.
Value
summary() returns a list with a pooled data frame, the
number of studies, estimation method and between-study covariance.
The table includes the number of studies reporting each cutoff.
predict() returns a data frame with accuracy and interval limits.
plot() invisibly returns the coordinates of the plotted points.
Details
The summary gives inverse logits of the fitted means and pointwise
Wald confidence intervals. The ROC plot places these fitted operating
points at one minus specificity and sensitivity. Multinomial predictions
evaluate the logistic distributions at the mean study locations. They
flag extrapolation beyond the observed cutoff range. Their pointwise
Wald intervals use the observed information when available and have
missing limits otherwise. For a latent-threshold summary curve from a
single-test bivariate fit, see dta_sroc().