Skip to main content
. Author manuscript; available in PMC: 2019 Oct 1.
Published in final edited form as: J Biomed Inform. 2018 Aug 30;86:149–159. doi: 10.1016/j.jbi.2018.08.014

Figure 4.

Figure 4.

This figure displays AUROC confidence intervals for each of 26 methodological combinations. AUROCs are ordered from top-to-bottom in descending order of AUROC from the expert-curated gold standard. The heatmap on the left indicates the presence (tan) or absence (blue) of each of the 6 method variables for each plotted AUROC. For example, the top AUROC method (according to the clinically-curated gold standard) used sequence-time, no binning, intra-patient normalization, differencing, no additional context variable, and a joint AR model. Note that these results are enumerated explicitly in Supplementary Table 1.