Abstract
Conventional reference ranges evolve from subjective criteria for health and disease. We offer an objective method for distinguishing typical from atypical values by purely statistical criteria. We define typical values as those exhibiting a linear relationship with percentiles on a value versus percentile plot. Identification of percentiles at which deviation from linearity occurs results from calculation of correlation coefficients between values and percentiles over centrally expanding ranges of percentiles. One selects arbitrarily some minimum value for these correlation coefficients, for example, 0.990, as the criterion for deviation from linearity. Values encompassed by these percentiles of deviation constitute an objective reference range. Identification of any correlations between atypical values and symptoms of disease requires clinical follow-up studies.
Collapse