Martin Koduah1, Terence C Iles, Barry J Nix. 1. School of Mathematics, Cardiff University, Senghennydd Road, PO Box 926, Cardiff CF24 4YH, United Kingdom.
Abstract
BACKGROUND: We introduce a new criterion, the percentile inclusion probability, for comparing methods for calculating reference intervals. The criterion is compared with a previously published measure of reliability suggested by Linnet (Linnet K. Clin Chem 1987;33:381-6), the ratio of the width of the confidence interval for the percentile to that of the reference interval. METHODS: Data were simulated from a range of theoretical statistical distributions representing the shapes of data sets encountered in clinical investigations. The two-stage transformation of the data to a gaussian distribution recommended by the IFCC was compared with a nonparametric approach. RESULTS: The percentile inclusion probability criterion identified that the parametric approach is in some cases seriously affected by bias. Using different parametric models, we compared nonparametric and parametric methods for two sets of clinical data and showed that the parametric approach is susceptible to model choice. CONCLUSIONS: Sample sizes significantly greater than those currently recommended are required to establish reference intervals, regardless of whether parametric or nonparametric methods are used. Parametric methods are preferable when the data are truly gaussian, but are only marginally better than nonparametric methods when data transformation is needed to achieve a gaussian shape.
BACKGROUND: We introduce a new criterion, the percentile inclusion probability, for comparing methods for calculating reference intervals. The criterion is compared with a previously published measure of reliability suggested by Linnet (Linnet K. Clin Chem 1987;33:381-6), the ratio of the width of the confidence interval for the percentile to that of the reference interval. METHODS: Data were simulated from a range of theoretical statistical distributions representing the shapes of data sets encountered in clinical investigations. The two-stage transformation of the data to a gaussian distribution recommended by the IFCC was compared with a nonparametric approach. RESULTS: The percentile inclusion probability criterion identified that the parametric approach is in some cases seriously affected by bias. Using different parametric models, we compared nonparametric and parametric methods for two sets of clinical data and showed that the parametric approach is susceptible to model choice. CONCLUSIONS: Sample sizes significantly greater than those currently recommended are required to establish reference intervals, regardless of whether parametric or nonparametric methods are used. Parametric methods are preferable when the data are truly gaussian, but are only marginally better than nonparametric methods when data transformation is needed to achieve a gaussian shape.