Selective participation and loss to follow-up in population-based longitudinal studies: novel approaches for dealing with unmeasured causes of selection