Thomas Lumley • 3/23/2018

Small p hacking

The article critiques the proposal to change statistical significance thresholds from p<0.05 to p<0.005. It argues the core problem is weak studies and flexible analysis choices, not the threshold itself. Using examples from genome-wide association studies (GWAS), it explains how very small p-values are sensitive to analytical decisions, especially in low-power scenarios.

0 comments

#data analysis #research methodology #statistics