mada (version 0.5.11)

phm-class: Methods for phm objects.

Description

Objects of the class phm are output by the function with the same name. Apart from standard methods the function sroc provides SROC curves and confidence bands for model fits.

Usage

# S3 method for phm
print(x, ...)
# S3 method for phm
summary(object, level = 0.95, ...)
# S3 method for phm
sroc(fit, fpr = 1:99/100, ...)
# S3 method for phm
plot(x, extrapolate = FALSE, confband = TRUE, level = 0.95,
     ylim = c(0,1), xlim = c(0,1), sroclty = 1, sroclwd = 1, 
     confbandlty = 2, confbandlwd = 0.5, ...)

Value

The sroc function returns a matrix ready for plotting. Each row corresponds to one point in ROC space.

Arguments

x

a phm object.

object

a phm object.

fit

a phm object.

level

numeric, the confidence level for calculations of confidence intervals (summary) or confidence bands (plot).

fpr

numeric, the false positives rates for which to calculate the predicted sensitivities.

extrapolate

logical, should the sroc curve be plotted beyond the observed false positive rates?

confband

logical, should confidence bands be plotted?

ylim

numeric of length 2, which section of the sensitivities to plot?

xlim

numeric of length 2, which section of the false positive rates to plot?

sroclty

integer, line type of the SROC curve

sroclwd

integer, line width of the SROC curve

confbandlty

integer, line type of the SROC curve's confidence band

confbandlwd

integer, line width of the SROC curve's confidence band

...

arguments to be passed on to other functions

Author

Philipp Doebler <[e-mail]>

Details

The SROC curve is derived from the model formula. The confidence bands are calculated from the bounds of the confidence interval for the diagnostic accuracy parameter \(\theta\). The parameter and its confidence interval are then also used to calculate the AUC and partial AUC using the formulae $$ AUC(a,b) = \int_a^bu^\theta\mathrm{d}u = \frac{1}{\theta+1}[b^{\theta+1}-a^{\theta+1}], $$ $$ AUC = AUC(0,1) $$ and $$ pAUC = \frac{1}{b-a}AUC(a,b), $$ where \(a\) is the lower bound of the observed false positive rates and \(b\) the upper.

References

Holling, H., Boehning D., Boehning, W. (2012) “Meta-Analysis of Diagnostic Studies based upon SROC-Curves: a Mixed Model Approach using a Proportional Hazards Model.” Statistical Modelling, 12, 347--375.

See Also

phm

Examples

# load data
data(AuditC)
# fit model
fit <- phm(AuditC)
#calculate a SROC curve, but do not plot it
sroc.AuditC <- sroc(fit)
# plot the SROC curve in ROC space as a line
plot(sroc.AuditC, type = "l")
# Fancy version using plot
plot(fit)