Thomas Lumley • 11/15/2013

Moving the goalposts?

The article analyzes a PNAS paper advocating for a stricter p-value threshold (0.005) to improve scientific reproducibility. The author critiques this approach, arguing it merely swaps false positives for false negatives, ignores the problem of weak data and small sample sizes, and fails to address the misuse of p-values as a publishing criterion. The core argument is that better experimental design, not just moving thresholds, is the real solution.

0 comments

#statistics #Hypothesis Testing #Bayesian Inference