Lesson 9: Investigating sleep: a correlational study and Pearson's r
Evaluate a media claim about eating late and sleep
SkillsCorrelation and Pearson's rDesigning an investigation
Worked example
A news article reports: "Midnight snacks wreck teen sleep. A survey of 200 teenagers found that those who ate later in the evening slept less (, ). Eating late causes poor sleep, so parents should ban food after 8 pm."
Evaluate the claim.
Step 1: What the data show
is a weak negative correlation: teenagers who ate later tended to sleep a little less. With 200 participants, , so the correlation is significant: unlikely to be chance.
Step 2: Strength in context
A weak correlation means the points are widely scattered: many late eaters sleep well and many early eaters sleep poorly. Eating time is related to only a small part of the variation in sleep.
Step 3: Cause
The survey is correlational: no one was assigned an eating time. The direction could be reversed (teenagers who stay up late, because of the adolescent shift in melatonin, get hungry and eat later), or a third variable such as late study, sport training, part-time work or screen use could cause both.
Step 4: The recommendation
Banning food after 8 pm is not justified by these data. An experiment, or at least controlling for bedtime and other variables, would be needed to test whether eating time affects sleep.
Step 5: Judgement
The finding of a significant but weak association is credible; the causal claim and the advice are not supported.
Answer
The data show a significant (p < 0.001) but weak negative correlation (r = −0.32). Because the study is correlational, it cannot show that eating late causes poor sleep: the direction may be reversed (late bedtimes lead to late eating) or a third variable (late study, work, screens) may cause both. The causal claim and the ban are not supported.
Try a similar one
A fresh question of the same kind, checked as you answer, with hints and a full solution.
Method
Describe direction, strength and significance accurately.
Judge whether the design can support a causal claim.
Offer a reverse direction and a third variable.
Make a clear judgement about the claim.
Watch out for
Treating "significant" as "strong". With a large sample, a weak correlation can be significant.