psp.univariate: Univariate estimation of the relative true positive fraction...

In mclements/psp: Some tools to work with paired screen-positive studies.

Description

Univariate estimation of the relative true positive fraction and relative false positive fraction from a paired screen-positive design. Variance estimation uses

Usage

psp.univariate(disease, data, test1, test0)

Arguments

disease	expression for the outcome
data	data-frame of the input data
test1	expression for the new test
test0	expression for the reference test

Details

This uses the variance estimator from Cheng and Macaluso (1997).

Value

returns a list of class psp.univ with elements rTPR,varlogrTPR,rFPR and varlogrFPR. At present, the only method available is print.psp.univ, which prints the point estimates and confidence intervals.

Examples

## Not run: 

## Prostate cancer example from Pepe and Alonzo (2001):

require(foreign)
ex <- read.dta("http://research.fhcrc.org/content/dam/stripe/diagnostic-biomarkers-statistical-center/files/psa_dre_v2.dta")

psp.univariate(disease=I(d=="yes"), data=ex,
               test1=I(psa=="pos"),
               test0=I(dre=="pos"))

## We can also do this, albeit less efficiently, using regression models:

summary(fit <- rTPR(I(d=="yes") ~ test,
                    data = ex,
                    test1=I(psa=="pos"),
                    test0=I(dre=="pos")))
summary(fit <- rFPR(I(d=="yes") ~ test,
                    data = ex,
                    test1=I(psa=="pos"),
                    test0=I(dre=="pos")))

## End(Not run)

mclements/psp documentation built on May 22, 2019, 3:10 p.m.