The value of the p-value
Why “p-value<0.05” is not enough
The interpretation of p-value is one of the most misunderstood subjects in science and statistics, not only by the general public but also for Academia. This leads to low-quality publications and misleading news, with a consequent distrust in Science. Just stating “p<0.05 = good” and “p-value≥0.05 = bad” is a dangerous oversimplification.
This article tries to debug some concepts, focusing on:
- Data distribution and description
- Hypothesis testing
- The value of the p-value
- Practical limitations of the p-value
- How to choose the test to calculate the p-value?
1. Data distribution and description
Let’s take, for example, the height in adult men vs women. Empirically, we all know that on average men are taller than women (we are talking about big numbers). But what if we did not know this and ask ourselves…
Is there a difference between the height of adult men vs adult women?
To answer this question objectively, we have to design a study to test a hypothesis.
