Time-dependent signal detection measures are needed when trial outcomes arrive dynamically

Evidence

The results above stay true if false alarms are taken into account (Fig. 1h ). For this purpose, we computed a time-dependent d’ measure (see Fig. 1g for an illustration, and Methods ). Classical d’ analysis does not apply in the present dynamic decision paradigm, since the eventual outcome of a trial is not accessible at early times (i.e. when computing d’(0.5 s), a trial with a hit at 5 s should not be counted as a hit, but as a ‘correct reject up to 0.5 s; but when computing d’(4 s), the same trial will be counted as a hit). — Dijk, Pim van, Physiology, Psychoacoustics and…, 3330

Categorization

• DEFINE: Explains why classical d prime fails in dynamic decision paradigms and how a time-dependent d-prime resolves the censoring problem.
• XREF: The methodological problem of right-censored outcomes resembles survival analysis and event-history statistics from other domains.