Colloquium, Trainings, and Events

  • Image for Colloquium on Statistical Sciences

    Colloquium on Statistical Sciences

    Last updated: May 26, 2021, 9:31 p.m.

    Speaker: Dr. Michael Daniel C. Lucagbo

    Date: June 4, 2021, 9 a.m.

    Venue: via Zoom

    Join us in our Colloquium on Statistical Sciences via Zoom! You may register here:

    The talks will also be livestreamed through Facebook Live at

    Rectangular Prediction Regions for Setting Reference Regions in Laboratory Medicine

    by Dr. Michael Daniel C. Lucagbo, UP School of Statistics and University of Maryland, Baltimore County

    Reference intervals are among the most widely used medical decision-making tools and are invaluable in the interpretation of laboratory test results of patients. Moreover, when there are multiple biochemical analytes measured from each patient, a multivariate reference region (MRR) is needed. Such regions are more desirable than separate univariate reference intervals, which disregard the cross-correlations among variables. Traditionally, in a multivariate normal setting, MRRs have been constructed as ellipsoidal regions. However, such regions cannot detect component-wise extreme values. Consequently, MRRs are rarely used in actual practice. To address this drawback of ellipsoidal reference regions, we propose procedures to construct rectangular MRRs. We use a prediction region criterion in the development of the procedures, and we evaluate their accuracies through estimated coverage probabilities and expected volumes. We also compare the procedures with Bonferroni simultaneous intervals and recommend a solution to incorporate covariates in a multivariate regression setting. Finally, we show applications to real-life studies, including the diagnosis of kidney function and construction of reference intervals for three components of the insulin-like growth factor (IGF) system.