sensitivity: Compute sensitivity/recall/true positive rate

In jpfrench81/neastbenchmark: Benchmark Data for Disease Clusters

Compute sensitivity/recall/true positive rate

Description

Compute the empirical sensitivity/recall/true positive rate of a method from a series of tests. In this context, the sensitivity is the average proportion of the population of the true hotspot that lies within the most likely cluster. The function requires the null test statistics, the results from the observed data sets (i.e., the maximum test statistic and most likely cluster from each data set), the true hotspot locations, and the vector of population sizes for each region. See Details.

Usage

sensitivity(tnull, tdata, hotspot, pop, alpha = c(0.05, 0.01))

recall(tnull, tdata, hotspot, pop, alpha = c(0.05, 0.01))

tpr(tnull, tdata, hotspot, pop, alpha = c(0.05, 0.01))

Arguments

tnull	The set of null test statistics
tdata	The list of maximum test statistics (tmax) and most likely cluster (mlc) for each simulated data set.
hotspot	A vector containing the hotspot indices for the current data set.
pop	A vector with the populations associated with each region.
alpha	The type I error rate. Default is c(0.05, 0.01).

Details

In this context, the sensitivity is the proportion of the true hotspot population lying within the most likely cluster, averaged over all tests. If the pop vector is a vector of 1s, then the sensitivity will be the average of the proportion of the most likely cluster regions intersecting the true hotspot, divided by the number of regions in the true hotspot.

Value

A vector of sensitivity values.

Examples

tnull = 1:99
tdata = list(list(tmax = 96, mlc = c(50, 51)),
             list(tmax = 101, mlc = c(48, 57)))
sensitivity(tnull, tdata, 50, pop = rep(1, 100))

jpfrench81/neastbenchmark documentation built on July 26, 2023, 3:07 p.m.