sampEstimate: Estimate the sample size required to achieve a specified...

In rplzzz/sampEstimator: Estimate Sample Sizes for Reliable Detection with an Imperfect Test

Description

Calculate the number of samples needed to estimate the population prevalence with a specified precision. Precision is defined in terms of the quantiles of the distribution of estimated prevalence values. For example, we might say, "If the true prevalence is 1 greater than 0.5 we don't grossly underestimate the prevalence.

Usage

sampEstimate(Npop, popprev, prob, targ, sensitivity, specificity)

Arguments

Npop	Population size
popprev	True population prevalence
prob	The quantile to place a restriction on (e.g., 0.05 for the 5th percentile)
targ	Target value for prob
sensitivity	Sensitivity of the test
specificity	Specificity of the test

Details

The characteristics of the system that are specified are:

• Total population size

• Test characteristics

• True population prevalence

• Quantile to place a requirement on

• Target value of that quantile

The solver actually solves for the sample size that is as close as possible to being equal to the required condition, so it is indifferent to whether we are setting an upper bound or a lower bound on the quantile. So, in place of the example above we could just as well say that we want the 95th percentile of estimated values to be less than 2

Solution is carried out using a bounded secant solver. No initial guess is required because we can start with the basic assumption that the number of samples must be greater than zero and less than the total population. Since the number of samples must be a whole number, we shrink the interval until it's less than a quarter-person wide, and then we return the ceiling of the upper bound.

rplzzz/sampEstimator documentation built on May 24, 2020, 4:35 a.m.