Is a neuron’s spiking response modulated by an event, such as visual stimulus presentation, optogenetic stimulation, or behavioral response? This question is central to many neuroscientific studies, but a best standard practice, curiously enough, does not exist. A common approach is to bin spike counts into trial-averages and perform a t-test across trial-repetitions to see if the response “during” is different from “before” a stimulus. This might be simple and often effective, but if the neuron shows a clear peak followed by a reduction in spiking, a t-test will not pick up that the neuron is responsive to the stimulus (see figure). Another approach is to make a peri-stimulus time histogram (PSTH) and perform an ANOVA across all bins. While this is somewhat better, the problem remains that the bin size is an arbitrary choice.
To solve these problems, Jorrit Montijn and colleagues at the Netherlands Institute for Neuroscience developed a new statistical method: the ZETA-test. This new method is binless and parameter-free, which allows researchers to avoid arbitrary parameter choices altogether. More importantly, it strongly outperforms alternatives, such as t-tests, ANOVAs, and model-based approaches, in the sense that the new method detects more stimulus-responsive neurons at the same level of false-positives.
This method could become an important new tool in neuroscience and other disciplines: the same amount of data will now provide more statistically robust results. Or alternatively, researchers could now use less data, and fewer experimental animals, to get the same statistical reliability.
Read “A parameter-free statistical test for neuronal responsiveness” here >>