Logo image
Back
Addressing bias when estimating pooled d’ for multiple cues or neurons
Abstract   Peer reviewed

Addressing bias when estimating pooled d’ for multiple cues or neurons

Daniel R. Guest
The Journal of the Acoustical Society of America, Vol.159(4_Supplement), pp.A151-A151
08/01/2026
DOI: 10.1121/10.0045375

View Online

Abstract

One topic of interest in sensory psychology and neuroscience is the sensitivity achieved by a hypothetical observer when task information is combined (or pooled) across multiple independent sources. For example, an observer might combine information provided by two independent sensory cues to make a judgement, or an experimenter might wish to characterize the sensitivity achieved by a population of independent neurons when pooling information across them. In the case of signal detection theory, the d’ of an observer optimally combining information across independent sources is equal to the square root of the sum of squared individual d’ values (optimal combination rule). However, naive application of the optimal combination rule to individual d’ estimates leads to biased estimates of pooled d’. Here, we demonstrate this phenomenon with computer simulations and characterize how its severity depends on parameters of interest, such as sample size and sensitivity. We suggest an alternative unbiased estimator for the optimal combination rule that could be useful to experimenters working with small sample sizes ( < ∼40 trials) and/or difficult tasks (d’ < 1), where the bias is most problematic. [Work supported by NIH R01 DC010813 and NIH F32 DC022143.]

Details

Metrics

1 Record Views
Logo image