Null Hypothesis
What the Null Hypothesis
Let’s use a simple example.
How Does a Blind Man Know It’s Raining?
Imagine a blind man walking outside. He can’t see a thing in front of him and can only rely on his sense of smell and touch.
While he’s walking outside, he starts to feel water droplets on his head. He thinks it might be raining, but he can’t say for sure because he can’t actually see the rain coming down from the sky.
In frequentist statistics, the hypothesis that says “it’s raining” is the alternative. The hypothesis that says there’s “there’s no rain” is the null hypothesis. The null hypothesis generally assumes that there’s normality (i.e., no rain).
Since the blind man can’t actually prove that there’s rain (from his lack of vision), he has to find evidence against the idea that there’s no rain.
That’s what statisticians do when they employ frequentist methods. They’re blind people who can’t actually prove that it’s raining. Instead, they provide evidence that whatever is happening is unlikely if it is not raining.
How does this work for the blind man? First, he knows he feels water drops on his head. That could be rain or it could be a sprinkler. However, he doesn’t hear a sprinkler. He does notice that the grass he’s walking through feels muddier than usual. That supports the idea that more water is coming from somewhere else. He also notices that as he walks, the water droplets are following him. If it were a sprinkler, the water probably wouldn’t follow him. Finally, he hears thunder in the distance.
Based on everything he feels and hears, it is very unlikely he’d experience all of that if it weren’t raining. Therefore, it’s likely that the water droplets are from rain.
That is the crux of how frequentist statistics works. You show that you wouldn’t have the results in the data that you have if nothing were happening at all. That allows you to reject the null in favor of the alternative. It does not disprove the null and does not allow you to prove the alternative.
“You’re making the assumption that the null hypothesis is true,”
The p-value he’s referring to is the probability of observing our results as or more extreme than what you observed given that the null hypothesis is true.
Let’s use a simple example.
How Does a Blind Man Know It’s Raining?
Imagine a blind man walking outside. He can’t see a thing in front of him and can only rely on his sense of smell and touch.
While he’s walking outside, he starts to feel water droplets on his head. He thinks it might be raining, but he can’t say for sure because he can’t actually see the rain coming down from the sky.
In frequentist statistics, the hypothesis that says “it’s raining” is the alternative. The hypothesis that says there’s “there’s no rain” is the null hypothesis. The null hypothesis generally assumes that there’s normality (i.e., no rain).
Since the blind man can’t actually prove that there’s rain (from his lack of vision), he has to find evidence against the idea that there’s no rain.
That’s what statisticians do when they employ frequentist methods. They’re blind people who can’t actually prove that it’s raining. Instead, they provide evidence that whatever is happening is unlikely if it is not raining.
How does this work for the blind man? First, he knows he feels water drops on his head. That could be rain or it could be a sprinkler. However, he doesn’t hear a sprinkler. He does notice that the grass he’s walking through feels muddier than usual. That supports the idea that more water is coming from somewhere else. He also notices that as he walks, the water droplets are following him. If it were a sprinkler, the water probably wouldn’t follow him. Finally, he hears thunder in the distance.
Based on everything he feels and hears, it is very unlikely he’d experience all of that if it weren’t raining. Therefore, it’s likely that the water droplets are from rain.
That is the crux of how frequentist statistics works. You show that you wouldn’t have the results in the data that you have if nothing were happening at all. That allows you to reject the null in favor of the alternative. It does not disprove the null and does not allow you to prove the alternative.
“You’re making the assumption that the null hypothesis is true,”
The p-value he’s referring to is the probability of observing our results as or more extreme than what you observed given that the null hypothesis is true.
Comments
Post a Comment