Your roommate Nadia has a heart condition, and her doctor says a rise of 5 beats a minute in her resting heart rate would be too much.
Nadia is sleeping badly and wants to try Dozeberry sleep gummies. The bottle says “Clinically tested. No effect on heart rate.”
No effect on heart rate is exactly what Nadia needs. Does the trial behind the label show no effect?
How the lab tested the claim
To check a claim like this, a lab runs a hypothesis test. The null hypothesis is the bottle’s claim, no effect: the population mean change in resting heart rate, for everyone who takes the gummies, is 0.
The lab can’t test everyone, so it uses a sample. It tracked 40 adults for a week without the gummies and a week taking one each night, and found each person’s change in resting heart rate.
In the first picture, each dot is one person’s change, from −19 to 24 beats a minute. The diamond is the sample mean of the 40 changes, a rise of 2.5.
The second picture zooms in on the sample mean, from −4 to 10 beats a minute.
The shaded band runs from −3.1 to 3.1. The line through the band’s center marks 0, the value in the null hypothesis. The band is where the sample mean can fall without the lab rejecting the null hypothesis.
A sample mean in the band is close enough to 0 that it would not be unusual if the population mean change were 0. The diamond is inside the band, so the lab does not reject the null hypothesis.
Does the trial show what the bottle says, no effect on heart rate?
You chose yes.
Not quite. The bottle’s claim, no effect, was the null hypothesis, and the lab didn’t reject it. But not rejecting a null hypothesis isn’t the same as showing that it’s true.
You chose no.
That’s right. The bottle’s claim, no effect, was the null hypothesis, and the lab didn’t reject it. But not rejecting a null hypothesis isn’t the same as showing that it’s true.
You said you’re not sure.
That’s all right. The bottle’s claim, no effect, was the null hypothesis, and the lab didn’t reject it. But not rejecting a null hypothesis isn’t the same as showing that it’s true.
What about a rise of 5?
Nadia’s doctor set the limit at 5 beats a minute. Can the trial rule out a population mean change of 5?
To find out, the null hypothesis can say the population mean change is 5, not 0.
The number line of changes, from −4 to 10, and the diamond at 2.5 stay the same. Only the band and its line move.
See whether the diamond stays inside the band: press 5, the doctor’s limit.
Null hypothesis: population mean change
The lab’s test doesn’t reject the null hypothesis that the population mean change is 5, either. With that null hypothesis, the band runs from 1.9 to 8.1, and the diamond is inside it.
The population mean change can’t be both 0 and 5.
What it adds up to
So does the trial show no effect on heart rate? No. There is not enough evidence to conclude that the population mean change differs from 0, but the trial can’t rule out a population mean change of 5.
If there is not sufficient evidence to reject the null hypothesis, we conclude that the null hypothesis might be true, but we never conclude that the null hypothesis is true.
