Interpreting a Correlation in the External Assessment
A study reports a correlation of r = -0.62 between hours of social media use per day and self-reported sleep quality in a sample of 400 adolescents. The researchers also report that adolescents who used social media more than 4 hours daily were twice as likely to report poor sleep. A newspaper headline states: "Social media ruins teenagers' sleep." A student writes in an exam that the negative correlation means social media use causes poor sleep, and that the study proves reducing screen time will improve sleep for all teenagers.
1. Interpret the correlation coefficient and evaluate the newspaper headline.
Read the worked answer
The coefficient r = -0.62 indicates a moderate-to-strong negative correlation: as reported daily social media use increases, reported sleep quality tends to decrease. The magnitude, 0.62, is substantial but not perfect, so a meaningful proportion of variance in sleep quality is not explained by social media use. The figure of twice the likelihood of poor sleep for heavy users describes an association in the sample, not a causal risk in the population.
The headline "Social media ruins teenagers' sleep" asserts causation, which a correlational design cannot support. There are at least three plausible explanations: social media use may disrupt sleep, poor sleep may lead to more social media use, or a third variable such as anxiety, loneliness, or evening light exposure may influence both. The headline also generalises from a sample to all teenagers. A defensible statement would be that heavier social media use is associated with poorer reported sleep quality among the adolescents studied, and that experimental or longitudinal evidence is needed before causal claims are made.
2. The student claims the study proves that reducing screen time will improve sleep for all teenagers. Identify the errors in this reasoning.
Read the worked answer
Causation from correlation: the design is correlational, so it cannot establish that social media use causes poor sleep. Without manipulation or control of confounders, reverse causation and third variables remain viable.
Overgeneralisation: the sample of 400 adolescents may not represent all teenagers, and the claim about "all teenagers" ignores individual differences in vulnerability, baseline sleep, and family context.
Prediction versus proof: even if an association exists, it does not prove that an intervention will work. An intervention could fail if the relationship is driven by a third variable such as anxiety, in which case reducing screen time alone would not improve sleep.
Measurement issues: sleep quality is self-reported, which may be biased and may correlate with mood or reporting style rather than objective sleep.
A more defensible conclusion is that the findings justify further investigation, ideally a randomised or longitudinal design measuring objective sleep, before recommending universal screen-time reduction as an intervention.