spider (version 1.5.1)

threshOpt: Threshold optimisation

Description

Determines the positive, negative, false positive and false negative rates of identification accuracy for a given threshold.

Usage

threshOpt(distobj, sppVector, threshold = 0.01)

Value

A table giving the threshold and number of negative and positive identifications, number of false negative and false positive identifications, and the cumulative error.

Arguments

distobj

Distance matrix.

sppVector

Species vector (see sppVector).

threshold

Threshold distance for delimiting intra- and inter-specific variation. Default of 0.01.

Author

Rupert Collins <[e-mail]>

Details

When run over a range of thresholds, this function allows the optimisation of threshold values based on minimising the identification error rates. See the example below for more details.

References

Meyer, C. P., and Paulay, G. (2005). DNA barcoding: error rates based on comprehensive sampling. _PLoS Biology_ *3* (12), 2229-2238.

See Also

localMinima.

Examples

data(anoteropsis)
anoDist <- ape::dist.dna(anoteropsis)
anoSpp <- sapply(strsplit(dimnames(anoteropsis)[[1]], split="_"), 
    function(x) paste(x[1], x[2], sep="_"))
threshOpt(anoDist, anoSpp)

data(dolomedes)
doloDist <- ape::dist.dna(dolomedes)
doloSpp <- substr(dimnames(dolomedes)[[1]], 1, 5)
threshOpt(doloDist, doloSpp)

#Conduct the analysis over a range of values to determine the optimum threshold
threshVal <- seq(0.001,0.02, by = 0.001)
opt <- lapply(threshVal, function(x) threshOpt(doloDist, doloSpp, thresh = x))
optMat <- do.call(rbind, opt)
graphics::barplot(t(optMat)[4:5,], names.arg=optMat[,1], xlab="Threshold values", 
    ylab="Cumulative error")
graphics::legend(x = 2.5, y = 29, legend = c("False positives", "False negatives"), 
    fill = c("grey75", "grey25"))