Hi Everyone,
I'm excited to announce that we have a guest speaker for next month. Radha Langoju will present:
When Correct Statistics Lead to Wrong Science: Understanding the Replication Crisis
Abstract: The replication crisis has exposed a paradox at the core of modern science: many results that are statistically correct are nevertheless scientifically false. Across disciplines, findings produced using standard statistical procedures increasingly fail to replicate—not due to errors or misconduct, but because statistical tools are often applied ritualistically rather than inferentially.
This talk argues that the crisis reflects a failure of reasoning, not mathematics. Through case studies such as the rise and collapse of “power posing,” it illustrates how low prior plausibility, flexible analytical choices, misinterpreted p-values, and model misspecification systematically inflate false discoveries. Structural incentives—including novelty-driven publication practices, underpowered studies, and the overuse of conceptual replication—further amplify unreliable results.
The presentation concludes by proposing a shift from ritual to rigor: prioritizing direct replication, emphasizing severity-based testing over significance thresholds, and adopting a balanced approach that gives equal weight to study design, measurement quality, and statistical analysis. Addressing the replication crisis requires systematic reform in scientific practice, not new statistical rituals.
About the Speaker: Radha Langoju is a Data Scientist with over six years of experience in data analytics, machine learning, and AI, spanning the academic, healthcare, and nonprofit sectors. Her work focuses on applied research, impact evaluation, and graduate-level training, with a particular emphasis on translating complex statistical and machine learning methods into actionable insights. She is currently on a break and volunteering with Statistics Without Borders.