Everyone Uses P-Values, But No One Knows What They Are

For indispensable reporting on the coronavirus crisis, the election, and more, subscribe to the Mother Jones Daily newsletter.


A blue-ribbon committee of the American Statistical Association spent a year arguing about what a p-value is, and finally coughed up the following explanation aimed at laymen:

Informally, a p-value is the probability under a specified statistical model that a statistical summary of the data (for example, the sample mean difference between two compared groups) would be equal to or more extreme than its observed value.

Raise your hand if you understood a word of that. Anyone?

Among experts, the long-running argument about p-values is abstruse and knotty. But even when you ignore the deep philosophical issues, it turns out that coming up with a close-enough explanation for average Joes and Janes is tough too. Over at 538, Christie Aschwanden hints at the problem: statisticians aren’t so much interested in explaining what a p-value is as they are in busting myths and explaining what it isn’t:

A common misconception among nonstatisticians is that p-values can tell you the probability that a result occurred by chance. This interpretation is dead wrong, but you see it again and again and again and again. The p-value only tells you something about the probability of seeing your results given a particular hypothetical explanation — it cannot tell you the probability that the results are true or whether they’re due to random chance. The ASA statement’s Principle No. 2: “P-values do not measure the probability that the studied hypothesis is true, or the probability that the data were produced by random chance alone.”

Personally, I’ve never been too happy with this approach, but I’ll leave that aside. I’m probably just wrong. But how about this nickel explanation?

If you’re testing a hypothesis with only a limited set of data (for example, proposing that someone is the leader of a presidential race by relying on a survey of only 1,000 people) a p-value is, informally, the probability that the small dataset validated your hypothesis merely by chance.

I suppose that’s wrong too in some kind of barely comprehensible way. It always is. But close! And, perhaps, reasonably comprehensible?

Dear Reader,

This feels like the most important fundraising drive since I've been CEO of Mother Jones, with staggeringly high stakes and so much uncertainty. In "News Is Just Like Waste Management," I try to unpack the reality we all face and how we can rise to the challenge. If you're able to, this is a critical moment to support Mother Jones’ nonprofit journalism: We need to raise $400,000 to help cover the vital reporting projects we have planned, and right now is no time to pull back.

Monika Bauerlein, CEO, Mother Jones

Dear Reader,

This feels like the most important fundraising drive since I've been CEO of Mother Jones, with staggeringly high stakes and so much uncertainty. In "News Is Just Like Waste Management," I try to unpack the reality we all face and how we can rise to the challenge. If you're able to, this is a critical moment to support Mother Jones’ nonprofit journalism: We need to raise $400,000 to help cover the vital reporting projects we have planned, and right now is no time to pull back.

Monika Bauerlein, CEO, Mother Jones

We Recommend

Latest

Sign up for our free newsletter

Subscribe to the Mother Jones Daily to have our top stories delivered directly to your inbox.

Get our award-winning magazine

Save big on a full year of investigations, ideas, and insights.

Subscribe

Support our journalism

Help Mother Jones' reporters dig deep with a tax-deductible donation.

Donate