A0202
Title: Random-covariate-dependent rectangular reference regions under multivariate normality
Authors: Raden Gerald Agustin - University of the Philippines Cebu (Philippines) [presenting]
Michael Daniel Lucagbo - University of the Philippines Diliman (Philippines)
Abstract: Reference intervals are among the most widely used decision-making tools in the medical field and are invaluable in interpreting laboratory test results. These intervals may depend on covariates such as age and sex. The covariate values are often random quantities since they are typically not controlled in reference interval determination studies. When several biochemical analytes are needed to diagnose the same condition, the use of combined univariate reference intervals is not recommendable since the analytes could be correlated. Instead, a multivariate reference region is necessary to consider the correlations among the analytes. Traditionally, multivariate reference regions have been constructed as ellipsoidal. However, such regions cannot detect the outlying ness of a specific analyte. Procedures are proposed to construct rectangular multivariate reference regions incorporating random covariate information from the subjects. The reference regions are computed in a multivariate normal setting, and the prediction region criterion is used. A parametric bootstrap approach is employed to calculate the required prediction factor. Numerical results show that the parametric bootstrap approach is entirely accurate, with coverage probabilities very close to the desired nominal value. Finally, the proposed method is applied to real-life data from a study to compute covariate-dependent reference regions for insulin-like growth factors.
