Linear correlation is insufficient as the sole measure of associations: The case of technology use and mental health.

It is common for psychology studies to rely solely on linear correlation (r) or similar statistics and not include other measures of association (such as relative risk, which examines differences in the number of people affected). For example, the association between smoking and lung cancer (r = 0.06) could be dismissed as "small" if only linear r is examined, even though 30 times more smokers than non-smokers get lung cancer. Many studies concluding that associations between technology use and well-being as too small to be of practical importance relied solely on linear r. We show that, across five datasets, "small" correlations between technology use and mental health exist alongside practically important risk associations. As there are several valid types of association, and characterizing an association based on a single type of a measure - such as linear r or r - can be misleading.