The problem of power in single-case neuropsychology
Case-control comparisons are a gold standard method for diagnosing and researching neuropsychological deficits and dissociations at the single-case level. These statistical tests, developed by John Crawford and collaborators, provide quantitative criteria for the classical concepts of deficit, dissociation and double-dissociation. Much attention has been given to the control of Type I (false positive) errors for these tests, but far less to the avoidance of Type II (false negative) errors; that is, to statistical power. I will describe the origins and limits of statistical power for case-control comparisons, showing that there are hard upper limits on power, which have important implications for the design and interpretation of single-case studies. My aim is to stimulate discussion of the inferential status of single-case neuropsychological evidence, particularly with respect to contemporary ideals of open science and study preregistration.
Panel discussion: Practical advice for reproducibility in neuroscience
This virtual, interactive panel on reproducibility in neuroscience will focus on practical advice that researchers at all career stages could implement to improve the reproducibility of their work, from power analyses and pre-registering reports to selecting statistical tests and data sharing. The event will comprise introductions of our speakers and how they came to be advocates for reproducibility in science, followed by a 25-minute discussion on reproducibility, including practical advice for researchers on how to improve their data collection, analysis, and reporting, and then 25 minutes of audience Q&A. In total, the event will last one hour and 15 minutes. Afterwards, some of the speakers will join us for an informal chat and Q&A reserved only for students/postdocs.