Faster replacements for calculating ROC and PR curve data than with
ROCR::prediction() and ROCR::performance()
compute_roc(yvec, pvec)compute_pr(yvec, pvec)
Replacements to use instead of a combination of ROCR::prediction()
and ROCR::performance() to calculate ROC and PR curves. These functions are
about 10 to 20 times faster when using mcmcRocPrc() with curves = TRUE
and/or fullsims = TRUE.
See this issue on GH (ShanaScogin/BayesPostEst#25) for more general details.
And here is a note with specific performance benchmarks, compared to the old approach relying on ROCR.