Performance Evaluation of Two Software for Analysis Through ROC Curves: Comp2ROC vs SPSS

摘要

Receiver Operating Characteristic (ROC) analysis is a powerful tool to evaluate, view and compare diagnostic tests by a discriminating way. Currently there several ROC analysis tools, but none is known, by containing all the features necessary for a full investigation. In this work we determined the performance of Comp2ROC package (R package) and its functionality by performing comparative diagnostic systems based on empirical ROC curves for unpaired and paired samples. We compare this package with the functionality of IBM~® SPSS~® Statistics to analyse ROC curve, in order to determine whether it has better ability to execute both the level of performance as a result. For illustrative purpose we use a random sample of clinical indexes used in neonatal intensive care to evaluate the risk of death for newborns with very low birth weight (VLBW) (＜ 1500g) and/or gestational age ＜ 32 weeks.