The Effective Statistician - in association with PSI P-value and confidence intervals - the good, the bad, and the ugly
Mar 10, 2025
In this insightful discussion, Kaspar Rufibach, an expert in statistical hypothesis testing, dives into the often-misunderstood world of p-values and confidence intervals. He explores the historical debates between Fisher’s and Neyman-Pearson’s approaches, shedding light on their practical implications. Kaspar shares real-world examples from clinical trials, emphasizing the risks of misinterpretation and the importance of clear communication. His tips for applying statistical concepts effectively can enhance collaboration with clinicians and strengthen the integrity of research.
AI Snips
Chapters
Transcript
Episode notes
Fisher's P-Value Approach
- Fisher's approach to p-values aimed to combine evidence from multiple experiments over time.
- This differs from making immediate decisions, allowing confident rejection of null hypotheses after sufficient evidence accumulation.
Neyman-Pearson Hypothesis Testing
- Neyman-Pearson's framework quantifies the risks of wrong decisions (Type I and II errors) in hypothesis testing.
- This framework allows explicit decisions for either the null or alternative hypothesis.
Statistical Significance is Binary
- Use "statistically significant" as a binary outcome in hypothesis testing, avoiding qualifiers like "highly significant".
- Remember that rejecting a null hypothesis is binary; it's either rejected or not.