Okay, so imagine you flip a coin 100 times and get heads 70 times. That’s weird, right?
A p-value tells you: “If this coin was actually fair, how weird would this result be?”
If the p-value is really small (like 0.001), it means: “Wow, this would almost never happen if the coin was fair.”
That’s basically it. It’s just a number that tells you how surprising your results are.
People mess this up a lot. So here’s what they’re definitely NOT:
Wrong ideas: - NOT “the chance I’m right” - NOT “the chance it happened by accident” - NOT “proof of anything”
What they actually are: - Just a number that measures how weird your data looks - A number based on assuming nothing special is happening
Think of it like a smoke detector. If it goes off, that’s surprising. But the alarm doesn’t tell you what’s on fire. It just says “something’s weird.”
Every test starts with two ideas fighting each other.
Null Hypothesis (H₀): “Nothing special is happening” - The medicine doesn’t work - The two groups are the same - The coin is fair
Alternative Hypothesis (H₁): “Something IS happening” - The medicine works - The two groups are different - The coin is rigged
The p-value helps you pick a winner. If the p-value is small enough, you say “okay, I believe H₁.”
\[p\text{-value} = P(\text{your result} \mid H_0 \text{ is true})\]
Translation: “What’s the probability of getting your result if nothing special is actually happening?”
For tests where extreme can go either direction:
\[p\text{-value} = P(|\text{result}| \geq |\text{observed}| \mid H_0)\]
You don’t need to memorize this. Just remember: small p-value = weird result = doubt the H₀
We asked 25 students how many hours they study per week:
study_hours <- c(5, 7, 6, 8, 4, 9, 5, 7, 6, 8, 5, 7, 6, 9, 5, 8, 6, 7, 5, 8, 6, 7, 5, 8, 7) t.test(study_hours, mu = 6)
P-value = 0.0548. Very close to our 0.05 threshold!
Red zone = results that would be pretty weird if H₀ was true
Before you run your test, pick a line: usually 0.05.
The rule: - p < 0.05: Result is weird. Reject H₀. - p ≥ 0.05: Result is normal. Keep H₀.
