Researchers have developed a novel method for calibrating statistical claims in large-scale hypothesis testing without requiring labeled data. This approach draws inspiration from probabilistic forecasting but adapts it to scenarios where the ground truth is never revealed, such as in multiple testing. By constructing pseudo-labels from ordered p-values, the method allows for stochastic assessment and indirect establishment of calibration, potentially improving the reliability of error probabilities, especially for measures like the q-value which were found to be severely miscalibrated in a survey of psychology and neuroscience literature. AI
IMPACT This research offers a new statistical framework for analyzing large datasets, potentially improving the reliability of findings in fields that heavily rely on hypothesis testing, including AI research.
RANK_REASON The cluster contains an academic paper detailing a new statistical methodology.
- arXiv
- Empirical Bayes with a changing prior
- False Discovery Rates Theory and Applications to DNA Microarrays
- Neuroscience
- Probabilistic forecasting
- Psychology
- P-Values
- Q value
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
