The reign of p-values is over. Statisticians rejoice!
March 27, 2016
In statistics we have some odd rule of thumbs, despite evidence that does little to support it. All our classic texts say that if we have a sufficient sample size then we can make assumptions about the distribution of the sample mean based on the Central Limit Theorem. In cases that involve the t-distribution, if n ≥ 30 then we can apply the CLT and use the z-value instead of the t-value, and yet the t-value for any given significance level is still pretty far from the respective z-value.
Similarly, the significance we place in the p-value and that a p-value < 0.05 indicates a significant result is a farcical thought for most statisticians, but is ingrained in us from the very first statistics course we take. This is further perpetuated by journals favouring submissions with "significant" results. This may all change, as the ASA has questioned the p-values importance and released a guideline on its use to improve the conduct and interpretation of quantitative science and inform the growing emphasis on reproducibility of science research. [Full article]