Association of disrupted body clock with mood disorders: 2 statistical problems
This cross-sectional study interests a psychologist for its small OR
However, these ORs, etc, represent associations between only a One-Quintile reduction in relative amplitude, a measure of
the extent to which circadian rhythmicity of rest–activity cycles is disrupted, and mood disorders, etc.
1/5 full body clock disruption ==?== mood disorders
UK residents aged 37–73 years were recruited into the UK Biobank general population cohort from 2006 to 2010. We used data from a subset of participants whose activity levels were recorded by wearing a wrist-worn accelerometer for 7 days.
From these data, we derived a circadian relative amplitude variable, which is a measure of the extent to which circadian rhythmicity of rest–activity cycles is disrupted.
In the same sample, we examined cross-sectional associations between low relative amplitude and mood disorder… A one-quintile reduction in relative amplitude was associated with …
In fact, the 2nd problem is really interesting. Some P-values are so small but corresponding ORs are not so extreme (all ≈ 1). That’s partly why some researchers propose abandoning p-value (threshold).