likelihood ratios (and associated CIs) from pROC

[yhpua]

I'm writing to ask if it is possible to add "positive and negative likelihood ratios" to the ret arg of ci.coords()?

Thank you for pROC and for considering my request!

[xrobin]

Thanks for the request, it should be feasible.
Do you have a reference for the exact formulas to calculate these 2 ratios?

[yhpua]

Thank you for your kind consideration.

To compute the CIs, the 2 oft-cited references are as follows:

Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763–770. doi: 10.1016/0895-4356(91)90128-V.

Martín-Andrés A, Álvarez-Hernández M. Two-tailed approximate confidence intervals for the ratio of proportions. Stat Comput. 2014;24:65–75. doi: 10.1007/s11222-012-9353-5.

[xrobin]

This was requested on SO: https://stackoverflow.com/q/77487146/333599
See Wikipedia for the formulas: https://en.wikipedia.org/wiki/Likelihood_ratios_in_diagnostic_testing

[xrobin]

The feature is now available in the develop branch:

> data(aSAH)
> roc.s100b <- roc(aSAH$outcome, aSAH$s100b)
> coords(roc.s100b, 0.2, ret=c("t", "se", "sp", "lr_pos", "lr_neg"))
          threshold sensitivity specificity   lr_pos    lr_neg
threshold       0.2   0.6341463   0.8055556 3.261324 0.4541632

> ci.coords(roc.s100b, 0.2, ret=c("lr_pos", "lr_neg"))
95% CI (2000 stratified bootstrap replicates):
    threshold lr_pos.low lr_pos.median lr_pos.high lr_neg.low lr_neg.median lr_neg.high
0.2       0.2      2.061         3.278       6.272     0.2773        0.4532      0.6548