Some effects of within-person variability in epidemiological studies