sensR (version 1.2.2)

AUC: AUC computation

Description

This is the default AUC function for scalar d-primes, which will compute Area Under the ROC curve (ROC is an acronym for receiver operating characteristic).

Usage

## S3 method for class 'default':
AUC(d, scale = 1, se.d, CI.alpha = 0.05, ...)
## S3 method for class 'discrim':
AUC(d, CI.alpha = 0.05, ...)

Arguments

d

a unit lenght vector with the value of d-prime for which AUC is to be computed or a discrim object from the fitting of a A-not A test with AnotA

scale

a unit length vector giving the ratio of scale (ie. standard deviation) of the latent distribution for the no-class items relative to that of the yes-class items

se.d

standard error of d (d-prime). If provided, the function will compute confidence limits of value of AUC---cf. in section value.

CI.alpha

the type I level of the confidence interval of AUC

...

additional arguments passed integrate

Value

• A list with components. If se.d is supplied to the default method or if a discrim object is supplied, the object contains the latter three additional elements.
• value
  the estimated value of AUC
• res.int
  the result from the call to integrate
• lower
  the lower confidence limit
• upper
  the upper confidence limit
• CI.alpha
  echoes the provided CI.alpha

Details

The function calls integrate to obtain the area under the ROC curve implied by d and scale. Confidence limits are based on a normal approximation of d and not of AUC. The limits are computed, if an estimate of the standard error of d is provided. Note that the limits do not take the uncertainty in estimating the scale nor that of estimating the standard error of d into account. A print method is implemented for objects of class AUC.

Examples

(odor <- matrix(c(112, 112, 72, 53, 22, 4, 7, 38, 50, 117, 101, 62), 2,
               byrow = TRUE))
(d.primes <- SDT(odor)[,3])
for(i in 1:5) print(AUC(d.primes[i]))
## Provide standard error of d-prime and compute CI:
fm1 <- AnotA(8, 25, 1, 25)
AUC(fm1$coef, , fm1$se)
AUC(fm1)