Is It Time to Ban the P Value?
- Type:#article
- Year read:#read2021
- Subject: Statistics | p-value
- Bibtex: @kraemer2019
- Bibliography: Kraemer, H. C. (2019). Is It Time to Ban the P Value? JAMA Psychiatry, 76(12), 1219–1220. https://doi.org/10.1001/jamapsychiatry.2019.1965
Why and when I was reading this
I was following up on the p-value debate.
Key takeaways
- Readers need to look beyond p-values and evaluate study quality and effect sizes instead
- P-values largely depend on sample size, so in very large studies everything will be significant
- P-values should be reported to a priori hypotheses
The reason people are criticizing the p-value is that it is misused, miscomputed, and misinterpreted.
A valid p-value has a distribution which is determined by sample size, reliability of sensitivity of the measures used, quality of the design and analyses, and quality of the research in general.
Even though a p-value might be well below 0.05 it may still lack clinical relevance, because in large studies even trivial differences yield low p-values.
In contrast, effect sizes remain the same regardless of sample size.
The author suggests a 3-part strategy.
Each p-value should be reported after a priori justification for its hypothesis
So a call for proper pre-registration of studies.
We shouldn’t do mass testing with p-values and pretend that all of those tests were created equal.
P-values should always be accompanied by descriptive statistics
Preferably, an effect size with confidence intervals.