Let’s talk about evidence-based medicine.
Suppose you’re in the hospital and a nurse takes your temperature to find out whether you have a fever. Providing that the thermometer is working properly, it will give you a number that answers the question. It’s all the evidence you need. It doesn’t matter how many other patients in the hospital have had their temperature taken lately.
But suppose you need to gather more sophisticated medical evidence: You want to know whether your version of a specific gene predisposes you to a certain disease. Let’s say scientists have just finished a study of people with that disease to see which gene variants turn up more often than usual.
Again, you get a numerical answer—not a temperature, but a “P value.” It’s the probability that even if there was no real link, the study would turn up the same number of people (or more) with your version of the gene. A low P value