d : Sensitivity at the optimal criterion location

Area of Science:

• Psychology
• Cognitive Science
• Psychophysics

Background:

• Signal detection theory (SDT) traditionally measures signal discriminability.
• Standard measure d$\backslash$ relies on strict assumptions (binormal distributions, equal variance/base rates).
• Existing alternatives (d_a, d$\backslash$) partially address violations of SDT assumptions.

Purpose of the Study:

• Introduce a new signal detection measure, d$\backslash$.
• Develop a measure robust to SDT assumption violations.
• Ground discriminability in a minimize error count (MEC) strategy.

Main Methods:

• Proposed a novel signal detection measure, d$\backslash$.
• Conducted simulations to compare d$\backslash$ with existing measures (d_a, d$\backslash$).
• Examined implications for bias metrics (β, c) at optimal criteria.

Main Results:

• d$\backslash$ is robust to violations of standard SDT assumptions.
• d$\backslash$ remains consistent across varying response biases.
• d$\backslash$ reflects discriminability changes related to base rates, unlike d_a.
• d$\backslash$ aligns with d$\backslash$ when optimizing for MEC but is criterion-independent.

Conclusions:

• d$\backslash$ provides a more robust and consistent measure of signal discriminability.
• The new measure is grounded in an observer's error minimization strategy.
• d$\backslash$ offers advantages over existing measures, particularly when SDT assumptions are violated.