Statistics of Longitudinal quality of life assessment in cancer trials: should data be treated as continuous or ordinal?